Methods and devices for processing data from attenuate sets formed from media data.

Description

1 / 87 “METHODS AND DEVICES FOR PROCESSING ATTRIBUTE SET DATA FORMED FROM MEDIA DATA

[0001] This application claims priority to U.S. Patent Application No. 18 / 643,559, filed April 23, 2024, and to U.S. Provisional Patent Application No. 63 / 498,201, filed April 25, 2023, and to U.S. Provisional Application No. 63 / 498,733, filed April 27, 2023, the entire contents of which are incorporated herein by reference. U.S. Patent Application No. 18 / 643,559, filed April 23, 2024, claims the benefit of U.S. Provisional Patent Application No. 63 / 498,201, filed April 25, 2023, and U.S. Provisional Patent Application No. 63 / 498,733, filed April 27, 2023. TECHNICAL FIELD

[0002] This disclosure relates to the transport of keypoint data extracted from video data. BACKGROUND

[0003] Digital video capabilities can be incorporated into a wide range of devices, including digital televisions, digital direct broadcast systems, wireless broadcast systems, personal digital assistants (PDAs), laptop or desktop computers, digital cameras, digital recording devices, digital media players, video game consoles, cellular or satellite radiotelephones, video conferencing devices, and the like. Digital video devices implement video compression techniques, such as those described in the standards defined by MPEG-2, MPEG-4, ITU-T H.263 or ITU-T H.264 / MPEG-4 Part 10, Advanced Video Coding (AVC Petition 870250092796, 10 / 10 / 2025, p. 9 / 207 2 / 87 Advanced Video Coding), ITU-T H.265 (also called High Efficiency Video Coding (HEVC) and extensions of these standards, to transmit and receive digital video information more efficiently.

[0004] Video compression techniques perform spatial and / or temporal prediction to reduce or remove the inherent redundancy in video sequences. For block-based video coding, a video frame or slice can be partitioned into macroblocks. Each macroblock can be further partitioned. Macroblocks in an intracoded frame or slice (I) are coded using spatial prediction relative to neighboring macroblocks. Macroblocks in an intercoded frame or slice (P or B) can use spatial prediction relative to neighboring macroblocks in the same frame or slice or temporal prediction relative to other reference frames.

[0005] After the video data has been encoded, the video data can be packaged (packetized) for transmission or storage. The video data can be assembled into a video file conforming to any of several standards, such as the International Organization for Standardization (ISO) base media file format and extensions thereof, such as AVC. SUMMARY

[0006] In general, this disclosure describes techniques for performing bitrate adaptation during streaming of attribute data over a network, such as a radio access network (RAN). Petition 870250092796, dated 10 / 10 / 2025, page 10 / 207 3 / 87 5G, inter-network tasking. For example, attributes can be extracted and processed from input video data. The attribute extraction and processing tasks can be divided and performed by different devices in different locations. Thus, a first device can extract attributes from input video data to form an attribute map and perform a first set of processes on the attribute map, form an intermediate attribute map, and then send the intermediate attribute map to a second device via a network, for example, 5G. The second device can then perform the remaining processing tasks on the intermediate attribute map.

[0007] According to the techniques of this disclosure, the first device may buffer video data from which the intermediate attribute map is formed and / or the intermediate attribute map itself. The second device may then determine whether the remaining processing tasks can be performed solely from the intermediate attribute map or whether the buffered media data / attribute map is required (or would benefit) the remaining processing tasks. If the buffered video data / attribute map is required or would provide a benefit, the second device may request the buffered data and, after receiving the buffered data, perform the remaining processing tasks using the intermediate attribute map and the retrieved buffered data (media data and / or attribute sets). For example, the source device may compress the set of Petition 870250092796, dated 10 / 10 / 2025, page 11 / 207 4 / 87 attributes to a reduced compression ratio and then send the reduced compression version of the attribute set to the destination device. Alternatively, the second device can request the buffered data only when sufficient network bandwidth is available between the first and second devices. This way, bandwidth consumption can be reduced by sending only the buffered data as needed.

[0008] In one example, a method for processing attribute set data formed from media data includes: performing, by a first network entity, a first part of a computation graph on a media dataset to form an attribute map, where the first part of the computation graph represents first processing tasks performed by the first network entity, and a second part of the computation graph corresponds to second processing tasks to be performed by a second network entity; buffering, by the first network entity, at least a portion of the media data; compressing, by the first network entity, the attribute map to form a compressed attribute map;and send, through the first entity, the compressed attribute map and additional data corresponding at least to the buffered portion of the media data to the second network entity to enable the second network entity to perform the second part of the computation graph using the attribute map and the additional data.

[0009] In another example, a device for processing attribute set data formed from Petition 870250092796, dated 10 / 10 / 2025, page 12 / 207 5 / 87 of media data includes: a memory configured to store attribute set data and media data; and a processing system comprising one or more processors implemented in a circuit array, wherein the processing system is configured to: perform a first part of a computation graph on a set of media data to form an attribute map, wherein the first part of the computation graph represents first processing tasks performed by the device, wherein the device comprises a first network entity, wherein a second part of the computation graph corresponds to second processing tasks to be performed by a second network entity; buffer at least a portion of the media data in memory; compress the attribute map to form a compressed attribute map;and send the compressed attribute map and additional data corresponding to at least the buffered portion of the media data to the second network entity to enable the second network entity to perform the second part of the computation graph using the attribute map and additional data.

[0010] In another example, a method for processing attribute set data formed from media data, wherein the method comprises: receiving, by a first network entity, a compressed attribute map including attributes extracted from media data and processed according to a first part of a computation graph and additional data from a second network entity, wherein the additional data corresponds at least to the portion of the media data that was stored in Petition 870250092796, dated 10 / 10 / 2025, page 13 / 207 6 / 87 buffer by the second network entity; decompress, by the first network entity, the compressed attribute map; and perform, by the first network entity, a second part of the computation graph on the attribute map using the additional data, where the second part of the computation graph corresponds to processing tasks to be performed by the first network entity.

[0011] In another example, a device for processing attribute set data formed from media data includes: a memory configured to store attribute set data and media data; and a processing system comprising one or more processors implemented in a circuit array, wherein the processing system is configured to: receive a compressed attribute map including attributes extracted from media data and processed according to a first part of a computation graph and additional data from a second network entity, wherein the additional data corresponds to at least a portion of the media data that was buffered by the second network entity; decompress the compressed attribute map;and perform a second part of the computation graph on the attribute map using the additional data, where the second part of the computation graph corresponds to processing tasks to be performed by the first network entity.

[0012] The details of one or more examples are set forth in the attached drawings and in the description below. Other attributes, objectives and advantages will become apparent from the description and drawings, and the claims. Petition 870250092796, dated 10 / 10 / 2025, page 14 / 207 7 / 87 BRIEF DESCRIPTION OF THE DRAWINGS

[0013] Figure 1 is a block diagram illustrating an exemplary system that implements techniques for streaming media data over a network.

[0014] Figure 2 is a conceptual diagram that illustrates an exemplary system that can perform the techniques of this disclosure.

[0015] Figure 3 is a graph representing the performance results of attribute compression for video encoding for machines.

[0016] Figure 4 is a graph illustrating an exemplary computer graph according to the techniques of this disclosure.

[0017] Figure 5 is a conceptual diagram illustrating a bar graph representing an output data size compared to a latency amount for various split points of a convolutional neural network (CNN) for image recognition.

[0018] Figure 6 is a block diagram that illustrates an exemplary system architecture for implementing the techniques of this disclosure.

[0019] Figure 7 is a call flow diagram that illustrates an example method that can be implemented by the system architecture of Figure 6 according to the techniques of this disclosure.

[0020] Figure 8 is a block diagram that illustrates an alternative system architecture for implementing the techniques described in this disclosure.

[0021] Figure 9 is a call flow diagram that illustrates an example method that can be Petition 870250092796, dated 10 / 10 / 2025, page 15 / 207 8 / 87 implemented using the system architecture shown in Figure 8, in accordance with the techniques described in this publication.

[0022] Figure 10 is a block diagram that illustrates an alternative system architecture for implementing the techniques described in this disclosure.

[0023] Figure 11 is a call flow diagram that illustrates an example method that can be implemented by the system architecture of Figure 10 according to the techniques of this disclosure. DETAILED DESCRIPTION

[0024] In general, this disclosure relates to techniques related to video coding for machines (VCM). VCM can be used when processing image or video data that is not necessarily intended for use by human users, for example, for video playback. For instance, video data can instead be processed for machine-based operational tasks such as security monitoring, object detection / tracking, instance segmentation, navigation for automated devices, machine vision, hybrid vision, or similar. In general, such processes can be performed using a task network, for example, using one or more artificial intelligence / machine learning (AI / ML) models. The task network can correspond to a computation graph, including nodes that represent various processing tasks to populate the task network.

[0025] The processing tasks of a task network can be divided at a split point, so that a first set of processing tasks Petition 870250092796, dated 10 / 10 / 2025, page 16 / 207 9 / 87 of the task network (that is, a first part of a computation graph representing the task network) is performed by a first network entity (e.g., a source device), and a second set of processing tasks of the task network is performed by a second network entity (e.g., a destination device). The source device may include a camera for capturing image or video data and a processing system configured to perform the first set of processing tasks. The source device may be, for example, a user equipment (UE) device. The source device may capture image or video data and perform the first set of processing tasks on the image or video data to extract a set of attributes, then compress the attribute set and send the compressed attribute set to the destination device.The target device can then unpack the set of attributes and perform the second set of processing tasks.

[0026] The first network entity can determine the split point, separating processing tasks into the first set of processing tasks and the second set of processing tasks, according to various operational conditions. For example, such conditions may include operations supported by the first network entity and the second network entity, bandwidth conditions of a network connection between the first network entity and the second network entity, the battery charge level of the source device / first network device, or similar conditions. The first Petition 870250092796, dated 10 / 10 / 2025, page 17 / 207 10 / 87 network entity can additionally determine encoding configuration information to be applied when encoding the attribute map, for example, based on bandwidth conditions and other operational conditions. In this way, processing tasks can be efficiently distributed while reducing bandwidth consumption when available bandwidth is reduced, thus improving the overall performance of the image processing system.

[0027] In some cases, the target device may perform the second set of processing tasks and determine that the results have a low confidence value. This may occur because, for example, the attribute map extracted from the media data during the first set of processing tasks may not include sufficient information, the decompressed attribute map may have lost much information during compression and decompression, or similar. Therefore, in some cases, the original media data (or some version thereof) may be needed to improve the confidence value of the second set of processing tasks. According to the techniques of this disclosure, the source device may buffer the media data (video or image data) used to form the attribute map, and the target device may request the buffered media data.The destination device can then receive the buffered media data from the source device and use the media data to perform the second set of processing tasks.

[0028] Additionally or alternatively, the Petition 870250092796, dated 10 / 10 / 2025, page 18 / 207 11 / 87 The source device may buffer attribute sets extracted from the media data. In some cases, the loss introduced by compression may lead to low confidence values ​​for object detection as performed by the target device. Additionally, the media data may include identifying representations of people or objects that a user may want to obscure (e.g., for privacy). Therefore, instead of requesting the media data directly, the target device may request the buffered attribute sets (or some portion of the data therefrom).

[0029] In still other examples, both media data and attribute sets can be buffered on the source device and one or both can be requested by the destination device.

[0030] In this way, the destination device can request the buffered media data or attribute sets only when needed, instead of continuously streaming the buffered media data and / or attribute sets from the source device to the destination device. Thus, these techniques can therefore reduce the network bandwidth consumed by sending the buffered media data or attribute sets (due to the fact that uncompressed or reduced media data / attribute sets only need to be sent as needed), while also maintaining high confidence values ​​for the overall set of processing tasks (the complete computation graph) performed jointly by the source and destination devices, due to the fact that the device Petition 870250092796, dated 10 / 10 / 2025, page 19 / 207 The 12 / 87 destination can use the media data and / or buffered attribute sets as needed.

[0031] Figure 1 is a block diagram illustrating an example system 10 that implements techniques for streaming media data over a network. In this example, system 10 includes content preparation device 20, server device 60, and client device 40. Client device 40 and server device 60 are communicatively coupled by network 74, which may comprise the Internet. In some examples, content preparation device 20 and server device 60 may also be coupled by network 74, or by another network, or they may be directly and communicatively coupled. In some examples, content preparation device 20 and server device 60 may comprise the same device.

[0032] The content preparation device 20, in the example of Figure 1, comprises an audio source 22 and a video source 24. The audio source 22 may comprise, for example, a microphone that produces electrical signals representative of the captured audio data to be encoded by the audio encoder 26. Alternatively, the audio source 22 may comprise a storage medium that stores previously recorded audio data, an audio data generator, such as a computerized synthesizer, or any other audio data source. The video source 24 may comprise a video camera that produces video data to be encoded by the video encoder 28, a storage medium encoded with previously recorded video data, a generation unit of Petition 870250092796, dated 10 / 10 / 2025, page 20 / 207 13 / 87 video data, such as a computer graphics source, or any other video data source. The content preparation device 20 is not necessarily communicatively coupled to the server device 60 in all examples, but may store multimedia content on a separate medium that is read by the server device 60.

[0033] Raw audio and video data may comprise analog or digital data. Analog data may be digitized before being encoded by audio encoder 26 and / or video encoder 28. Audio source 22 may obtain audio data from a speaking participant while the speaking participant is speaking, and video source 24 may simultaneously obtain video data from the speaking participant. In other examples, audio source 22 may comprise a computer-readable storage medium comprising stored audio data, and video source 24 may comprise a computer-readable storage medium comprising stored video data. In this manner, the techniques described in this disclosure may be applied to live, streaming, real-time audio and video data or to archived and pre-recorded audio and video data.

[0034] Audio frames that correspond to video frames are, in general, audio frames that contain audio data that was captured (or generated) by audio source 22 contemporaneously with video data captured (or generated) by video source 24 that is contained within the video frames. For example, although a speaking participant generally produces audio data per speech, audio source 22 captures the audio data and video source 24 Petition 870250092796, dated 10 / 10 / 2025, page 21 / 207 14 / 87 captures the video data of the speaking participant at the same time, that is, while audio source 22 is capturing audio data. In this way, an audio frame can correspond temporally to one or more particular video frames. Consequently, an audio frame corresponding to a video frame generally corresponds to a situation in which audio and video data were captured at the same time, and for which an audio frame and a video frame comprise, respectively, the audio and video data that were captured at the same time.

[0035] In some examples, the audio encoder 26 may encode a timestamp in each encoded audio frame that represents a time at which the audio data for the encoded audio frame was recorded, and similarly, the video encoder 28 may encode a timestamp in each encoded video frame that represents a time at which the video data for an encoded video frame was recorded. In such examples, an audio frame corresponding to a video frame may comprise an audio frame comprising a timestamp and a video frame comprising the same timestamp. The content preparation device 20 may include an internal clock from which the audio encoder 26 and / or the video encoder 28 may generate the timestamps, or which the audio source 22 and the video source 24 may use to associate audio and video data, respectively, with a timestamp.

[0036] In some examples, audio source 22 may send data to audio encoder 26 that correspond to a time when the audio data was Petition 870250092796, dated 10 / 10 / 2025, page 22 / 207 15 / 87 recorded, and the video source 24 can send data to the video encoder 28 that corresponds to a time at which the video data was recorded. In some examples, the audio encoder 26 can encode a sequence identifier in encoded audio data to indicate a relative temporal order of encoded audio data, but without necessarily indicating an absolute time at which the audio data was recorded, and similarly, the video encoder 28 can also use sequence identifiers to indicate a relative temporal order of encoded video data. Similarly, in some examples, a sequence identifier can be mapped or otherwise correlated to a timestamp.

[0037] Audio encoder 26 generally produces a stream of encoded audio data, while video encoder 28 produces a stream of encoded video data. Each individual data stream (whether audio or video) can be called an elementary stream. An elementary stream is a single, digitally encoded (possibly compressed) component of a media presentation. For example, the encoded video or audio portion of the media presentation can be an elementary stream. An elementary stream can be converted into a packetized elementary stream (PES) before being encapsulated in a video file. In the same media presentation, a stream ID can be used to distinguish the PES packets belonging to an elementary stream from one another. The basic data unit of an elementary stream is a packetized elementary stream (PES) packet. Thus, encoded video data generally corresponds to video streams. Petition 870250092796, dated 10 / 10 / 2025, p. 23 / 207 16 / 87 elementary. Similarly, audio data corresponds to one or more of the respective elementary streams.

[0038] In the example in Figure 1, the encapsulation unit 30 of the content preparation device 20 receives elementary streams comprising encoded video data from the video encoder 28 and elementary streams comprising encoded audio data from the audio encoder 26. In some examples, each of the video encoder 28 and the audio encoder 26 may include packers to form PES packets from encoded data. In other examples, each of the video encoder 28 and the audio encoder 26 may interface with their respective packers to form PES packets from encoded data. In still other examples, the encapsulation unit 30 may include packers to form PES packets from both encoded audio and video data.

[0039] Video encoder 28 can encode video data from multimedia content in various ways to produce different representations of the multimedia content at various bitrates and with various characteristics, such as pixel resolutions, frame rates, compliance with various encoding standards, compliance with various profiles and / or profile levels for various encoding standards, representations that have one or multiple views (e.g., for two-dimensional or three-dimensional playback), or other characteristics. A representation, as used in this disclosure, may comprise one of audio data, video data, text data (e.g., for subtitles), or other data. The representation may include an elementary stream, such as a stream Petition 870250092796, dated 10 / 10 / 2025, page 24 / 207 17 / 87 of elementary audio or an elementary video stream. Each PES packet can include a stream_id that identifies the elementary stream to which the PES packet belongs. The encapsulation unit 30 is responsible for assembling elementary streams into streamable media data.

[0040] The encapsulation unit 30 receives PES packets for elementary streams of a media presentation from the audio encoder 26 and the video encoder 28 and forms corresponding Network Abstraction Layer (NAL) units from the PES packets. The encoded video segments can be organized into NAL units, which provide network-friendly video representation addressing applications such as video telephony, storage, broadcast, or streaming. NAL units can be categorized into Video Coding Layer (VCL) NAL units and non-VCL NAL units. VCL units may contain the core compression engine and may include block, macroblock, and / or slice-level data. Other NAL units may be non-VCL NAL units.In some examples, an image encoded in a time instance, typically presented as a primary encoded image, may be contained within an access unit, which may include one or more NAL units.

[0041] Non-VCL NAL units may include parameter set NAL units and SEI NAL units, among others. Parameter sets may contain sequence-level header information (in sequence parameter sets (SPS)) and the infrequently altered image-level header information (in image sets). Petition 870250092796, dated 10 / 10 / 2025, p. 25 / 207 18 / 87 of picture parameter sets (PPS). With parameter sets (e.g., PPS and SPS), rarely changing information does not need to be repeated for each sequence or image; thus, encoding efficiency can be improved. Furthermore, the use of parameter sets can enable out-of-band transmission of important header information, avoiding the need for redundant transmissions for error resilience. In out-of-band transmission examples, parameter set NAL units can be transmitted on a different channel than other NAL units, such as SEI NAL units.

[0042] Supplemental Enhancement Information (SEI) may contain information that is not necessary for decoding encoded image samples from VCL NAL units, but may assist in processes related to decoding, display, error resilience, and other purposes. SEI messages may be contained in non-VCL NAL units. SEI messages are the normative part of some standard specifications and are therefore not always mandatory for standard-compliant decoder implementation. SEI messages may be sequence-level SEI messages or image-level SEI messages. Some sequence-level information may be contained in SEI messages, such as resizability information SEI messages in the SVC example and view resizability information SEI messages in MVC.These example SEI messages can carry information about, for example, extraction. Petition 870250092796, dated 10 / 10 / 2025, page 26 / 207 19 / 87 operating points and the characteristics of the operating points.

[0043] Server device 60 includes a Real-time Transport Protocol (RTP) transmission unit 70 and a network interface 72. In some instances, server device 60 may include a plurality of network interfaces. In addition, any or all of the attributes of server device 60 may be implemented in other devices in a content delivery network, such as routers, bridges, proxy devices, switches, or other devices. In some instances, the intermediate devices in a content delivery network may cache multimedia content data 64 and include components that substantially adapt to those of server device 60. Generally, network interface 72 is configured to send and receive data via network 74.

[0044] The RTP transmission unit 70 is configured to deliver media data to client device 40 via network 74 in accordance with RTP, which is standardized in Request for Comment (RFC) 3550 by the Internet Engineering Task Force (IETF). The RTP transmission unit 70 may also implement RTP-related protocols such as the RTP Control Protocol (RTCP), the Real-time Streaming Protocol (RTSP), the Session Initiation Protocol (SIP), and / or the Session Description Protocol (SDP). The RTP transmission unit 70 may send media data via network interface 72, which may implement the protocol of Petition 870250092796, dated 10 / 10 / 2025, page 27 / 207 20 / 87 Uniform Datagram Protocol (UDP) and / or Internet Protocol (IP). Thus, in some examples, server device 60 can send media data via RTP and RTSP through UDP using network 74.

[0045] RTP transmission unit 70 can receive an RTSP description request, for example, from client device 40. The RTSP description request can include data indicating which data types are supported by client device 40. RTP transmission unit 70 can respond to client device 40 with data indicating media streams, such as media content 64, that can be sent to client device 40, along with a corresponding network location identifier, such as a Uniform Resource Locator (URL) or a Uniform Resource Name (URN).

[0046] The RTP transmission unit 70 can then receive an RTSP configuration request from client device 40. The RTSP configuration request can generally indicate how a media stream should be transported. The RTSP configuration request may contain the network location identifier for the requested media data (e.g., media content 64) and a transport specifier, such as local ports to receive RTP data and control data (e.g., RTCP data) on client device 40. The RTP transmission unit 70 can respond to the RTSP configuration request with an acknowledgment and data representing ports on server device 60 through which the RTP data and control data will be sent. The unit of Petition 870250092796, dated 10 / 10 / 2025, page 28 / 207 21 / 87 RTP transmission unit 70 can then receive an RTSP playback request, causing the media stream to be played, i.e., sent to client device 40 via network 74. The RTP transmission unit 70 can also receive an RTSP disassembly request to end the streaming session, in response to which the RTP transmission unit 70 can stop sending media data to client device 40 for the corresponding session.

[0047] The RTP receiving unit 52 can similarly initiate a media stream by initially sending an RTSP description request to the server device 60. The RTSP description request can indicate the data types supported by the client device 40. The RTP receiving unit 52 can then receive a response from the server device 60 specifying available media streams, such as media content 64, that can be sent to the client device 40, along with a corresponding network location identifier, such as a Uniform Resource Locator (URL) or a Uniform Resource Name (URN).

[0048] The RTP receiving unit 52 can then generate an RTSP configuration request and send the RTSP configuration request to the server device 60. As mentioned above, the RTSP configuration request can contain the network location identifier for the requested media data (e.g., media content 64) and a transport specifier, such as local ports to receive RTP data and control data (e.g., RTCP data) on the client device 40. In response, the RTP receiving unit 52 can Petition 870250092796, dated 10 / 10 / 2025, page 29 / 207 22 / 87 receive confirmation from server device 60, including the ports on server device 60 that server device 60 will use to send media data and control data.

[0049] After establishing a media streaming session between server device 60 and client device 40, the RTP transmission unit 70 of server device 60 can send media data (e.g., media data packets) to client device 40 according to the media streaming session. Server device 60 and client device 40 can exchange control data (e.g., RTCP data) indicating, for example, reception statistics by client device 40, so that server device 60 can perform congestion control or otherwise diagnose and resolve transmission faults.

[0050] Network interface 54 can receive and provide media from a selected media presentation to the RTP receiving unit 52, which can, in turn, provide the media data to the decapsulation unit 50. The decapsulation unit 50 can decapsulate elements of a video file into constituent PES streams, unpack the PES streams to recover encoded data, and send the encoded data either to the audio decoder 46 or to the video decoder 48, depending on whether the encoded data is part of an audio or video stream, for example, as indicated by the PES packet headers of the stream. The audio decoder 46 decodes encoded audio data and sends the decoded audio data to the audio output 42, while the video decoder 48 decodes encoded video data and sends the data to Petition 870250092796, dated 10 / 10 / 2025, page 30 / 207 23 / 87 decoded video, which may include multiple views of a stream, for video output 44.

[0051] Each of the video encoder 28, the video decoder 48, the audio encoder 26, the audio decoder 46, the encapsulation unit 30, the RTP reception unit 52 and the decapsulation unit 50 may be implemented as any one of several suitable sets of processing circuits, as applicable, such as one or more microprocessors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), discrete logic circuit sets, software, hardware, firmware or any combination thereof.Each of the video encoder 28 and video decoder 48 may be included in one or more encoders or decoders, each of which may be integrated as part of a combined video encoder / decoder (CODEC encoder / decoder). Similarly, each of the audio encoder 26 and audio decoder 46 may be included in one or more encoders or decoders, wherein each may be integrated as part of a combined CODEC. An apparatus comprising the video encoder 28, the video decoder 48, the audio encoder 26, the audio decoder 46, the encapsulation unit 30, the RTP reception unit 52 and / or the decapsulation unit 50 may comprise an integrated circuit, a microprocessor and / or a wireless communication device, such as a mobile phone. Petition 870250092796, dated 10 / 10 / 2025, page 31 / 207 24 / 87

[0052] Client device 40, server device 60, and / or content preparation device 20 may be configured to operate in accordance with the techniques in this disclosure. For example purposes, this disclosure describes these techniques with respect to client device 40 and server device 60. However, it should be understood that content preparation device 20 may be configured to perform these techniques instead of (or in addition to) server device 60.

[0053] The encapsulation unit 30 can form NAL units comprising a header that identifies a program to which the NAL unit belongs, as well as a payload, for example, audio data, video data, or data describing the transport or program stream to which the NAL unit corresponds. For example, in H.264 / AVC, a NAL unit includes a 1-byte header and a variable-size payload. A NAL unit that includes video data in its payload can comprise various levels of video data granularity. For example, a NAL unit can comprise a video data block, a plurality of blocks, a video data slice, or an entire video data image. The encapsulation unit 30 can receive encoded video data from the video encoder 28 in the form of PES packets of elementary streams. The encapsulation unit 30 can associate each elementary stream with a corresponding program.

[0054] The encapsulation unit 30 can also assemble access units from a plurality of NAL units. In general, an access unit can comprise one or more NAL units to represent a Petition 870250092796, dated 10 / 10 / 2025, page 32 / 207 25 / 87 video data frame, as well as audio data corresponding to the frame when such audio data is available. An access unit generally includes all NAL units for an output time instance, for example, all audio and video data for a time instance. For example, if each view has a frame rate of 20 frames per second (fps), then each time instance can correspond to a time interval of 0.05 seconds. During this time interval, the specific frames for all views of the same access unit (the same time instance) can be simultaneously rendered. In one example, an access unit might comprise an image encoded in a time instance, which can be presented as a primary encoded image.

[0055] Consequently, an access unit can comprise all audio and video frames of a common temporal instance, for example, all views corresponding to time X. This disclosure also refers to an encoded image of a particular view, as a view component. That is, a view component can comprise an image (or frame) encoded for a particular view at a particular time. Consequently, an access unit can be defined by comprising all view components of a common temporal instance. The decoding order of access units does not necessarily have to be the same as the output or display order.

[0056] After encapsulation unit 30 has NAL units and / or access units mounted on a video file based on the received data, encapsulation unit 30 passes the video file to the interface. Petition 870250092796, dated 10 / 10 / 2025, page 33 / 207 26 / 87 output 32 to the output. In some examples, the encapsulation unit 30 may store the video file locally or send the video file to a remote server via output interface 32, instead of sending the video file directly to the client device 40. Output interface 32 may comprise, for example, a transmitter, a transceiver, a device for recording data on a computer-readable medium, such as an optical drive, a magnetic media drive (e.g., floppy disk drive), a universal serial bus (USB) port, a network interface, or another output interface. Output interface 32 outputs the video file to a computer-readable medium, such as a broadcast signal, a magnetic medium, an optical medium, memory, a flash drive, or another computer-readable medium.

[0057] Network interface 54 can receive a NAL unit or a network access unit 74 and provide the NAL unit or the access unit to the decapsulation unit 50, via the RTP receiving unit 52. The decapsulation unit 50 can decapsulate elements of a video file into constituent PES streams, unpack the PES streams to recover encoded data, and send the encoded data either to the audio decoder 46 or to the video decoder 48, depending on whether the encoded data is part of an audio or video stream, for example, as indicated by the PES packet headers of the stream. The audio decoder 46 decodes encoded audio data and sends the decoded audio data to the audio output 42, while the video decoder 48 decodes encoded video data and sends the data to Petition 870250092796, dated 10 / 10 / 2025, page 34 / 207 27 / 87 decoded video, which may include multiple views of a stream, for video output 44.

[0058] Figure 2 is a conceptual diagram illustrating an exemplary system 100 that can implement the techniques of this disclosure. In particular, these techniques may include feature compression for video coding for machines (FC-VCM), for example, according to Updates on Video Coding for Machines, WG 2, MPEG Technical Requirements, July 2022, available at www.mpeg.org / wpcontent / uploads / mpeg_meetings / 139_OnLine / w21826.zip. As shown in Figure 2, system 100 includes task network part 104, an encoder 106, a decoder 108, and task network part 210. Together, task networks parts 1 (104) and 2 (110) form a complete task network, for example, a neural network, for example, segmentation, object tracking, object detection, or similar.

[0059] Task network part 104 extracts a set of attributes from the original image or video data 102 and provides an intermediate set of attributes (also called an attribute map, for example, the output of an intermediate stage of a computer graph (e.g., a neural network, a transformer, an R-CNN FPN)) to the encoder 106, which encodes the attribute data and sends the encoded attribute data to the decoder 108. The decoder 108 decodes the attribute data and provides the decoded attribute data to task network part 2 110. Task network part 2 completes the processing of the attribute data to Petition 870250092796, dated 10 / 10 / 2025, page 35 / 207 28 / 87 determine the results of task 112. In the example in Figure 2, system 100 also includes bitrate results 114, representing data determined according to the evaluation of a bitstream including the encoded attribute data.

[0060] In general, VCM is intended for consumption by a machine running an algorithm rather than a human. The task network formed by parts 1 and 2 of the task networks in Figure 2 can be for multiple tasks (e.g., implemented algorithms), such as object tracking or instance segmentation. If the task network is divided correctly, FC-VCM can reduce the bandwidth required (bit rate) for a communication system.

[0061] To perform VCM, task network part 104 can extract video attributes from the original image / video data 102. After performing one or more processing tasks on the extracted attributes, the encoder 106 can encode the current attributes and transmit the encoded attributes through a communication network. Compression can lead to performance loss in the task network, compared to the case where a first network entity (such as a UE) transmits the image or video through the communication network to a second network entity (e.g., an application server) that performs the entirety of the task network.

[0062] If the portion of the task network in the second network entity (including task network part 110) detects an object or has low confidence in the result and inference (as part of task results 112), the second network entity may request that the Petition 870250092796, dated 10 / 10 / 2025, page 36 / 207 29 / 87 The first network entity to send the image or video is for higher accuracy. For certain task networks, such as surveillance networks, there may be a long period of time before an object is detected. This request approach can keep the data rate low in the long term while improving task network performance.

[0063] According to the techniques of this disclosure, a first network entity, such as a user appliance (UE), may include components to perform task network part 1 104 and the encoder 106 to encode (compress) an intermediate attribute set (or intermediate attribute map) generated by task network part 1. The first network entity may send the defined encoded / compressed intermediate resource to a second network entity, such as an application server, which includes the decoder 108 and components for task network part 2 110. The first network entity may perform task network part 1 (e.g., an inference task network) and send compressed attributes to the second network entity and may buffer past images or videos. The buffered images or video frames may be identified, for example, by sequence numbers or timestamps.Buffered images or video frames can be compressed to reduce storage space. The amount of buffered data can be measured by the number of images or video frames, or the amount of time represented by the images or video frames.

[0064] The second network entity can Petition 870250092796, dated 10 / 10 / 2025, page 37 / 207 30 / 87 request that the first network entity send additional data for further inference. The additional data may include, for example, buffered images or video frames (which may be compressed), compressed attributes corresponding to some of the images or video frames at the same split point with less compression (which may improve inference performance), or attribute enhancement data (e.g., information lost in the initial compression), and / or compressed attributes corresponding to some images or video frames at a different split point, for example, a split point that results in improved inference performance compared to the initial split point.

[0065] The second network entity can construct the request to identify buffered media data (e.g., images or video frames), for example, using sequence numbers, frame numbers, picture order count (POC) values, or a playback time interval. The second network entity may additionally or alternatively specify, in the request, a target compression rate and / or a target bit rate for the requested data.The second network entity can send the request in response to any or all of the following: detection of an object by task network part 2, confidence in an inference result (e.g., measured against a threshold) of task network part 2, estimated communication network quality, predicted communication network quality, computing capacity of the first network entity, and / or latency requirements of task network part 2 (e.g., for object detection). A. Petition 870250092796, dated 10 / 10 / 2025, page 38 / 207 31 / 87 The second network entity can specify the cause and priority of the request to the first network entity, which can then prioritize the transmission of the requested additional data. For example, if the cause is the detection of a specific object, it may be urgent for the second network entity to receive the additional data to verify the detection or reduce the occurrence of false alarms.

[0066] The first network entity can send the requested additional data to the second network entity. The first network entity can send the additional data from the buffer directly or can generate the additional data (for example, by processing the buffered data differently or additionally) in response to the request. The first network entity can compress the additional data, for example, according to the requested compression rate or target bit rate, which may be different from the degree of compression applied to the initial attribute map.

[0067] The second network entity can then perform additional inference processing using the additional data received. For example, the second network entity can execute the same task network (part 2 of the task network) using the additional data. As another example, the second network entity can execute a different task network using the additional data, for example, at a different split point, using additional tasks or a smaller number of tasks, or similar.

[0068] The second network entity may also request that the first network entity adjust (for example, increase or decrease) an amount of data. Petition 870250092796, dated 10 / 10 / 2025, page 39 / 207 32 / 87 media is buffered based on the needs of performing additional inference processing. The amount of buffer can be modified in terms of a new number of images or video frames or attributes to be buffered and / or the time period to be covered by the buffered images, video frames, or attributes. Thus, the first network device can adjust the amount of buffered media data in response to this request.

[0069] The overall computation graph can initially be trained using system 100, in which the bitrate results 114 can be analyzed to determine a bitrate reduction resulting from the encoding / compression performed by the test encoder 106. The test decoder 108 can decode the intermediate attribute set, task network part 2 can perform the second set of processing tasks, and task results 112 can be compared with target data for input video / image data 102 to determine confidence values ​​for training the overall network.Once the training has been completed, a deployed system similar to system 100 can be used, although, in such a deployed system, the bit rate results 114 do not necessarily need to be generated, and the test encoder 106 and test decoder 108 can be replaced by real encoders and decoders, for example, as discussed in more detail below in relation to Figures 6, 8 and 10.

[0070] Figure 3 is a graph representing the performance results of attribute compression for Petition 870250092796, dated 10 / 10 / 2025, page 40 / 207 33 / 87 Video encoding for machines. The performance of the task network can depend on the bit rate. In the graph in Figure 3, the mean average precision (MAP) is compared with the bit rate per pixel (BPP), for example, segmentation tasks. As shown, the FC-VCM curve 120 indicates that FC-VCM offers significant performance gain compared to, for example, the attribute anchoring performance indicated by the attribute anchoring curve 124 and image anchoring techniques indicated by the image anchoring curve 122. FC-VCM also offers privacy and computation offloading, due to the fact that a person or other entity cannot detect identifying characteristics from an attribute map.

[0071] In general, the bitstream bit rate should match the communication network capacity. The communication network capacity available to an encoder may change over time, for example, due to user mobility and network load. A realistic communication network should be considered. Factors affecting the attribute bit rate may include the division and selection of internal task network structures, attribute quantization, and / or attribute compression. This disclosure describes techniques related to signaling aspects between the two endpoints for FC-VCM that can enable dynamic divisions and / or selection of the internal structure of a compute graph in response to bitrate adaptation for FC-VCM. This may include representation and / or negotiation.

[0072] Figure 4 is a graph illustrating an exemplary computation graph 130 according to the Petition 870250092796, dated 10 / 10 / 2025, page 41 / 207 34 / 87 techniques of this disclosure. In some examples, several split points and potential disposable internal structures can be determined. The example in Figure 4 represents internal structures including convolution unit (conv) 352, ReLU unit 354, MaxPool unit 360, convolution unit 364, ReLU unit 368, convolution unit 374, convolution unit 376, global average clustering unit 382, ​​and SoftMax unit 386.

[0073] In this example, the convolution unit 352 receives data_0 350. The convolution unit 352 can have a W of <64 x 3 x 3 x 3> and a B of <64> The convolution unit 352 can emit conv1_1 354, and the ReLU unit 356 can process conv1_1 354 to form conv1_2 358. The MaxPool unit 360 can process conv1_2 358 to form pool1_1 362. The convolution unit 364 can have a W of <16 x 64 x 1 x 1> and a B of <16> The convolution unit 364 can process pool1_1 362 to form fire2 / squeeze1x1_1 366.

[0074] The ReLU unit 368 can process fire2 / squeeze1x1_1 366 to form one or both of fire2 / squeeze1x1_2 370 and / or fire2 / squeeze1x1_2 372. The convolution unit 374 can have a W of <64 x 16 x 1 x 1> and a B of <64> and can process fire2 / squeeze1x1_2 370 to form fire2 / expand1x1_1 378. The convolution unit 376 can have a W of <64 x 16 x 3 x 3> and a B of <64> and can process fire2 / squeeze1x1_2 372 to form fire2 / expand3x3_1 380. Additional units not shown in Figure 4 may be part of the task network. Ultimately, the global medium clustering unit 382 can receive the processed data and form pool10_1 384. The SoftMax unit 386 can process pool10_1 384 to form softamxout_1 388. Petition 870250092796, dated 10 / 10 / 2025, page 42 / 207 35 / 87

[0075] A task network repository can include data representing various identifiers of potential split points and / or potential disposable internal structures of a computation graph, such as the one shown in Figure 4. The computation graph can be, for example, MLP, CNN, LSTM, GRU, transformer, or similar. The task network repository can send this data to the two entities involved in splitting the task network, for example, a user appliance (UE) and an application server (AS). The AS can perform part 1 of the task network and the encoder-side functionality, while the UE can perform the decoder functionality and part 2 of the task network, for example.

[0076] A split point can be the output of a compute node in the compute graph. For example, the split point can be the output of a MAX POOL2 layer of an AlexNet. In the example in Figure 4, any of data_0 350, conv1_1 354, conv1_2 358, pool1_1 362, fire2 / squeeze1x1_1 366, fire2 / squeeze1x1_2 370, fire2 / squeeze1x1_2 372, fire2 / expand1x1_1 378, fire2 / expand3x3_1 380, pool10_1 384, or softmaxout_1 388 can correspond to the split point between task network part 1 and task network part 2.

[0077] An internal structure to be discarded can be a computing node or the output of a computing node. For example, P2 in an R-based convolutional neural network (CNN) FPN network can be discarded.

[0078] Identifiers for split points and / or disposable internal structures can be markers generated by applying a computer graph description tool to the computer graph, Petition 870250092796, dated 10 / 10 / 2025, page 43 / 207 36 / 87 which can also be flagged. For example, the markers circled in Figure 4 represent examples of such identifiers. Identifiers can be encoded markers (to reduce size / bit rate). For example, the encoding can output 8 bits if there are 256 possible combined potential split points and potential internal structures that can be discarded. Two sets of identifiers can be used: one for potential split points and another for potential internal structures to be discarded. The network repository can be placed with an entity, for example, the application server.

[0079] The task network repository may also submit the compute graph description tool or an indication of the compute graph description tool. Examples of compute graph description tools include Open Neural Network Exchange (ONNX) and Neural Network Exchange Format (NNEF). In some instances, the task network repository may alternatively submit a URI, URL, or URN that defines and marks the split points or internal structures to be discarded.

[0080] Figure 5 is a conceptual diagram illustrating a 140-bar graph representing an output data size compared to a latency amount for various split points of a convolutional neural network (CNN) for image recognition.

[0081] Figure 6 is a block diagram illustrating an exemplary system architecture 150 for implementing the techniques of this disclosure. According to the example Petition 870250092796, dated 10 / 10 / 2025, p. 44 / 207 37 / 87 of Figure 6, a first network entity (first device 160) can send additional data in the form of buffered images or video frames (which can be compressed) to a second network entity (second device 170), for example, in response to a request for the buffered images or video frames from the second device 170. As shown in the example in Figure 6, the first device 160 (a first network entity) includes task network part 162, attribute compression unit 164, and buffer 166 for storing m media data frames. Task network part 162 and attribute compression unit 164 can be implemented in hardware, software, firmware, or any combination thereof (and when implemented in software or firmware, an indispensable set of processing circuits and storage media for storing instructions to be executed can be provided).

[0082] Buffer 166 can refer to any storage medium (one or more storage devices) that can store image or video data until it is later requested (or discarded). Buffer 166 can be configured to operate as a first-in, first-out (FIFO) buffer, so that once buffer 166 is full, if a new frame is received, the oldest received frame still in the buffer can be discarded to allow storage of the new frame.

[0083] The second device 170 (a second network entity) includes the attribute decompression unit 172, the task network part 2 174, the logic for Petition 870250092796, dated 10 / 10 / 2025, p. 45 / 207 38 / 87 request buffered frames 176 and memory 178. Task network part 2 174, attribute decompression unit 172, and logic for requesting buffered frames 176 can be implemented in hardware, software, firmware, or any combination thereof (and when implemented in software or firmware, an indispensable set of processing circuits and storage media for storing instructions to be executed can be provided). Memory 178 can refer to any storage media (one or more storage devices) for storing attribute set / map data and media data. The logic for requesting buffered frames 176 can request buffered frames from the first device 160 and store the received frames in memory 178.The attribute decompression unit 172 can also store decompressed attribute sets in memory 178, and the task network part 2 174 can retrieve the decompressed attribute sets and / or media data from memory 178 when performing the second set of computing graph processing tasks.

[0084] The first device 160 and the second device 170 communicate through the communication network 152.

[0085] As discussed above, the logic for requesting buffered frames 176 from the second device 170 may request additional data based on, for example, one or more of the following: detection of an object, confidence in an inference result, estimated quality of the represented communication network, predicted quality of the communication network, computing capacity of the first device. Petition 870250092796, dated 10 / 10 / 2025, page 46 / 207 39 / 87 device and / or a latency requirement of a particular attribute processing task. In this example, the additional data includes the images or frames buffered 166 by the first device 160, for example, one or more of the n, n-1,... nm-1 frames. That is, the logic for requesting the buffered frames 176 (which may be implemented in hardware, software, firmware, or a combination thereof) may request one or more of the buffered frames and / or request that the size of buffer 166 be adjusted. In response, the first device 160 may send the requested buffered frames and / or adjust the size of buffer 166, for example, to store more or less than m frames / images.

[0086] In this way, the first device 160 represents an example of a device for processing attribute set data formed from media data, including: a memory configured to store attribute set data and media data; and a processing system comprising one or more processors implemented in a circuit array, wherein the processing system is configured to: perform a first part of a computation graph on a set of media data to form an attribute map, wherein the first part of the computation graph represents first processing tasks performed by the device, wherein the device comprises a first network entity, wherein a second part of the computation graph corresponds to second processing tasks to be performed by a second network entity; buffer at least a portion of the media data in memory; compact the map Petition 870250092796, dated 10 / 10 / 2025, page 47 / 207 40 / 87 of attributes to form a compressed attribute map; and send the compressed attribute map and additional data corresponding at least to the buffered portion of the media data to the second network entity to enable the second network entity to perform the second part of the computation graph using the attribute map and the additional data.

[0087] Similarly, the second device 170 represents an example of a device for processing attribute set data formed from media data, including: a memory configured to store attribute set data and media data; and a processing system comprising one or more processors implemented in a circuit array, wherein the processing system is configured to: receive a compressed attribute map including attributes extracted from media data and processed according to a first part of a computation graph and additional data from a second network entity, wherein the additional data corresponds to at least a portion of the media data that was buffered by the second network entity; decompress the compressed attribute map;and perform a second part of the computation graph on the attribute map using the additional data, where the second part of the computation graph corresponds to processing tasks to be performed by the first network entity.

[0088] Figure 7 is a call flow diagram illustrating an example method that can be implemented by the system architecture of Figure 6 according to the techniques of this disclosure. The method in Figure 7 can Petition 870250092796, dated 10 / 10 / 2025, page 48 / 207 41 / 87 to be performed by the first device 160 and the second device 170 of Figure 6.

[0089] Initially, the first device 160 (e.g., a UE) can perform task network part 1 (180). That is, the first device 160 can extract a set of attributes from one or more media data images by performing a first part of a computation graph (a first set of processing tasks of a task network). Task network part 162 can process the media data images according to the first portion of the task network to form an intermediate attribute set. The resource compression unit 164 can encode (compress) the intermediate attribute set, and the first device 160 can then send the compressed attribute set to the second device 170 (182). Furthermore, the first device 160 can buffer the media data images in buffer 166.Buffered images may encompass at least a round-trip time (RTT) between the first device 160 and the second device 170 and the computation time of task network part 2 174. In some examples, the RTT and computation time may be the minimum amount of time to elapse before the first device 160 receives a request from the second device 170.

[0090] The attribute decompression unit 172 of the second device 170 can decode (decompress) the intermediate attribute set. Then, task network part 2 174 can perform task network part 2 (184). Based on the analysis of the task results, the second device 170 can request frames Petition 870250092796, dated 10 / 10 / 2025, page 49 / 207 42 / 87 stored in the buffer of the first device 160 (186). For example, the second device 170 may detect a particular object from the attribute set or determine that an object is present, but be unable to identify the object with a high confidence value. In response to the request, the first device 160 may send one or more images or video data frames from buffer 166 to the second device 170, according to the request (188).

[0091] The second device 170 can use the additional data (one or more images or video data frames from buffer 166, in this example) to perform additional inference processing (190), for example, to try to identify the object in the attribute set with a higher confidence value using the retrieved media data and the attribute set. In this example, the second device 170 also requests that the first device 160 adjust the size of buffer 166 (192). Furthermore, in some examples (such as if there were not enough buffered images or the number of buffered images exceeded what was needed to increase the confidence value), the second device 170, for example, increases or decreases the buffer size based on the results of the additional inference processing. In response, the first device 160 can adjust the size of buffer 166 according to the request of the second device 170 (194).

[0092] In this way, the method in Figure 7 represents an example of a data processing method for a set of attributes formed from media data that includes: performing, by a first network entity (for example, first device 160 in Figure 6), a first Petition 870250092796, dated 10 / 10 / 2025, page 50 / 207 43 / 87 part of a computation graph on a media dataset to form an attribute map, wherein the first part of the computation graph represents initial processing tasks performed by the first network entity, wherein a second part of the computation graph corresponds to secondary processing tasks to be performed by a second network entity; buffering, by the first network entity, at least a portion of the media data; compressing, by the first network entity, the attribute map to form a compressed attribute map; and sending, by the first entity, the compressed attribute map and additional data corresponding to at least the buffered portion of the media data to the second network entity to enable the second network entity to perform the second part of the computation graph using the attribute map and the additional data.

[0093] The method in Figure 7 also represents an example of a data processing method for a set of attributes formed from media data that includes: receiving, by a first network entity (e.g., second device 170 in Figure 6), a compressed attribute map including attributes extracted from media data and processed according to a first part of a computation graph and additional data from a second network entity, where the additional data corresponds at least to the portion of the media data that was buffered by the second network entity; decompressing, by the first network entity, the compressed attribute map; and performing, by the first network entity, a second part of the computation graph on the map. Petition 870250092796, dated 10 / 10 / 2025, page 51 / 207 44 / 87 of attributes using the additional data, where the second part of the computation graph corresponds to processing tasks to be performed by the first network entity.

[0094] Figure 8 is a block diagram illustrating another exemplary system architecture 200 for implementing the techniques of this disclosure. According to the example in Figure 8, a first network entity (first device 210) can send additional data in the form of compressed attributes stored in a buffer 216 that were processed by the same network portion of tasks (i.e., up to the same split point), but with less compression. Additionally or alternatively, the additional data can be enhancement data of the attributes, for example, information that was lost due to the initial compression.

[0095] As shown in the example in Figure 8, the first device 210 (a first network entity) includes task network part 1 212, a resource compression unit 214, and a buffer 216 for storing attribute sets for n, n-1,... nm-1 frames. Task network part 1 212 and the attribute compression unit 214 can be implemented in hardware, software, firmware, or any combination thereof (and when implemented in software or firmware, an indispensable set of processing circuits and storage media for storing instructions to be executed can be provided).

[0096] Buffer 216 can refer to any storage medium (one or more storage devices) that can store image or video data. Petition 870250092796, dated 10 / 10 / 2025, page 52 / 207 45 / 87 until they are requested later (or discarded). Buffer 216 can be configured to operate as a first-in, first-out (FIFO) buffer, so that once buffer 216 is full, if a new set of attributes for a frame is received, the oldest received set of attributes still in the buffer can be discarded to allow storage of the new set of attributes.

[0097] The second device 220 (second network entity) includes the attribute decompression unit 222, the task network part 2 224, the logic for requesting buffered attributes 226, and the memory 228. The task network part 2 224, the attribute decompression unit 222, and the logic for requesting buffered attributes 226 may be implemented in hardware, software, firmware, or any combination thereof (and when implemented in software or firmware, an indispensable set of processing circuits and storage media for storing instructions to be executed may be provided). Memory 228 may refer to any storage media (one or more storage devices) for storing attribute set / map data.The logic for requesting buffered attributes 226 can request buffered attributes from the first device 210 and store the received attribute sets in memory 228. The attribute decompression unit 222 can also store decompressed attribute sets in memory 228, and the task network part 2 224 can retrieve the decompressed attribute sets and / or attribute sets. Petition 870250092796, dated 10 / 10 / 2025, p. 53 / 207 46 / 87 retrieved from memory 228 when performing the second set of computer graphics processing tasks.

[0098] The first device 210 and the second device 220 communicate through the communication network 202, which may represent the Internet.

[0099] As discussed above, the logic for requesting attributes stored in buffer 226 of the second device 220 may request additional data based on, for example, one or more of the following: object detection, confidence in an inference result, estimated quality of the represented communication network, predicted quality of the communication network, computing capacity of the first device, and / or a latency requirement of a particular attribute processing task. In this example, the additional data includes attributes for the previously processed images or frames, where the attributes may be stored in buffer 216 of the first device 210, for example, attributes for one or more of the n, n-1,... nm-1 frames. The attributes of the additional data may be less compressed than the intermediate attribute set initially sent to the second device 220.The logic for requesting attributes stored in buffer 226 (which may be implemented in hardware, software, firmware, or a combination thereof) may request attributes for one or more of the buffered frames and / or request that the size of buffer 216 be adjusted. In response, the first device 210 may send the requested buffered attributes to frames and / or adjust the size of buffer 216, for example, to store attributes for more or fewer than m frames / images. Petition 870250092796, dated 10 / 10 / 2025, page 54 / 207 47 / 87

[00100] In this way, the first device 210 represents an example of a device for processing attribute set data formed from media data, including: a memory configured to store attribute set data and media data; and a processing system comprising one or more processors implemented in a circuit array, wherein the processing system is configured to: perform a first part of a computation graph on a set of media data to form an attribute map, wherein the first part of the computation graph represents first processing tasks performed by the device, wherein the device comprises a first network entity, wherein a second part of the computation graph corresponds to second processing tasks to be performed by a second network entity; buffer at least a portion of the media data in memory;compress the attribute map to form a compressed attribute map; and send the compressed attribute map and additional data corresponding to at least the buffered portion of the media data to the second network entity to enable the second network entity to perform the second part of the computation graph using the attribute map and the additional data.

[00101] Similarly, the second device 220 represents an example of a device for processing attribute set data formed from media data, including: a memory configured to store attribute set data and media data; and a processing system comprising one or more Petition 870250092796, dated 10 / 10 / 2025, page 55 / 207 48 / 87 processors implemented in a circuit array, wherein the processing system is configured to: receive a compressed attribute map including attributes extracted from media data and processed according to a first part of a computation graph and additional data from a second network entity, wherein the additional data corresponds to at least a portion of the media data that was buffered by the second network entity; decompress the compressed attribute map; and perform a second part of the computation graph on the attribute map using the additional data, wherein the second part of the computation graph corresponds to processing tasks to be performed by the first network entity.

[00102] Figure 9 is a call flow diagram that illustrates an example method that can be performed by the system architecture of Figure 8 according to the techniques of this disclosure. The method in Figure 9 can be performed by the first device 210 and the second device 220 of Figure 8.

[00103] Initially, the first device 210 (e.g., a UE) can perform task network part 1 (230). That is, the first device 210 can extract a set of attributes from one or more media data images. Task network part 1 212 can process the set of attributes according to a first portion of a task network, to form an intermediate set of attributes. The resource compression unit 214 can buffer the attributes for a certain number of frames and encode the intermediate set of attributes, and the first device 210 can then send the set of Petition 870250092796, dated 10 / 10 / 2025, p. 56 / 207 49 / 87 attributes packed into the second device 220 (232). The first device 210 can also buffer the set of attributes in buffer 216.

[00104] The attribute decompression unit 222 of the second device 220 can decode the intermediate attribute set. Then, task network part 224 can perform the second set of processing tasks (a second part of the computation graph) (234). Based on the analysis of the task results, the second device 220 can request buffered attributes for the frames from the first device 210 (236). For example, the second device 220 can detect a particular object from the attribute set. In response, the first device 210 can send attributes for one or more images or video data frames from the buffer 216 to the second device 220, according to the request (238). In particular, the attributes for one or more images or frames can be encoded with less loss (e.g., at a higher bit rate) than the intermediate attribute set initially sent to the second device 220.

[00105] The second device 220 can use the additional data (the attributes extracted from one or more images or video data frames from buffer 216, in this example) to perform additional inference processing (240). In this example, the second device 220 also requests that the first device 210 adjust the size of buffer 216 (242), for example, to increase or decrease the buffer size based on the results of the additional inference processing. In response, the first device 210 can adjust the size of buffer 216 accordingly. Petition 870250092796, dated 10 / 10 / 2025, page 57 / 207 50 / 87 request for the second device 220 (244).

[00106] In this way, the method in Figure 9 represents an example of a data processing method for a set of attributes formed from media data that includes: performing, by a first network entity (for example, first device 210 in Figure 8), a first part of a computation graph on a set of media data to form an attribute map, where the first part of the computation graph represents first processing tasks performed by the first network entity, and a second part of the computation graph corresponds to second processing tasks to be performed by a second network entity; buffering, by the first network entity, at least a portion of the media data; compressing, by the first network entity, the attribute map to form a compressed attribute map;and send, through the first entity, the compressed attribute map and additional data corresponding at least to the buffered portion of the media data to the second network entity to enable the second network entity to perform the second part of the computation graph using the attribute map and the additional data.

[00107] The method in Figure 9 also represents an example of a data processing method for a set of attributes formed from media data that includes: receiving, by a first network entity (e.g., second device 220 in Figure 8), a compressed attribute map including attributes extracted from media data and processed according to a first part of a computation graph and additional data from a Petition 870250092796, dated 10 / 10 / 2025, page 58 / 207 51 / 87 second network entity, wherein the additional data corresponds at least to the portion of the media data that was buffered by the second network entity; decompress, by the first network entity, the compressed attribute map; and perform, by the first network entity, a second part of the computation graph on the attribute map using the additional data, wherein the second part of the computation graph corresponds to processing tasks to be performed by the first network entity.

[00108] Figure 10 is a block diagram illustrating another system architecture for implementing the techniques of this disclosure. According to the example in Figure 10, a first network entity can send additional data in the form of compressed attributes corresponding to some images or video frames, but extracted using a different part of a task network, for example, ending with a different split point that can improve inference performance compared to the initial split point, or using additional tasks or a smaller number of tasks from the task network.

[00109] As shown in the example in Figure 6, the first device 270 (a first network entity) includes task network part 1 272, attribute compression unit 274, buffer 276 for storing attributes for m media data frames extracted according to an original task network part 1 up to a first split point, and buffer 278 for storing attributes for the m media data frames up to a second split point (e.g., a new task network part 1 that is only Petition 870250092796, dated 10 / 10 / 2025, page 59 / 207 52 / 87 part of the original task network part 1 or includes the original task network part 1). Task network part 1 272 and attribute compression unit 274 may be implemented in hardware, software, firmware, or any combination thereof (and when implemented in software or firmware, an indispensable set of processing circuits and storage means for storing instructions to be executed may be provided).

[00110] Buffers 276 and 278 can refer to any storage medium (one or more storage devices) that can store image or video data and / or attribute data until later requested (or discarded). Buffers 276 and 278 can be configured to operate as a first-in, first-out (FIFO) buffer, so that once buffer 276 / 278 is full, if a new data object (image, video frame, or attribute set) is received, the oldest received data object still in the buffer can be discarded to allow storage of the new data object.

[00111] In this example, the second device 280 (a second network entity) includes the attribute decompression unit 282, the task network part 2 284, and the logic for requesting buffered attributes and / or frames 286. Memory 288 can refer to any storage media (one or more storage devices) for storing attribute set / map data and media data. The logic for requesting buffered attributes and / or frames 286 can request buffered attributes and / or frames from the first device 270 and store the attributes and / or frames. Petition 870250092796, dated 10 / 10 / 2025, page 60 / 207 53 / 87 received in memory 288. The attribute decompression unit 282 can also store decompressed attribute sets in memory 288, and the task network part 2 284 can retrieve the decompressed attribute sets and / or media data from memory 288 when performing the second set of computing graph processing tasks.

[00112] The first device 270 and the second device 280 communicate through the communication network 252, which may correspond to the Internet.

[00113] As discussed above, the logic for requesting attributes and / or frames stored in buffer 286 of the second device 280 may request additional data based on, for example, one or more of the following: object detection, confidence in an inference result, estimated quality of the represented communication network, predicted quality of the communication network, computing capacity of the first device, and / or a latency requirement of a particular attribute processing task. In this example, the additional data may include images from buffer 276 and / or attributes for image frames stored in buffer 278 of the first device 270 that were extracted and / or processed according to a split point that is different from the original split point associated with buffer 276.

[00114] The attributes can be for one or more of the n, n-1, ... nm-1 frames. That is, the logic for requesting the frames stored in buffer 286 (which can be implemented in hardware, software, firmware, or a combination thereof) can request the attributes for one or more of the buffered frames and / or request that the size of buffer 276 and / or the size of buffer 278 be Petition 870250092796, dated 10 / 10 / 2025, page 61 / 207 54 / 87 adjusted. In response, the first device 270 can send the requested frames and / or the buffered attributes for the frames and / or adjust the buffer size 276 and / or the buffer size 278, for example, to store attributes for more or less than m frames / images.

[00115] In this way, the first device 270 represents an example of a device for processing attribute set data formed from media data, including: a memory configured to store attribute set data and media data; and a processing system comprising one or more processors implemented in a circuit array, wherein the processing system is configured to: perform a first part of a computation graph on a set of media data to form an attribute map, wherein the first part of the computation graph represents first processing tasks performed by the device, wherein the device comprises a first network entity, wherein a second part of the computation graph corresponds to second processing tasks to be performed by a second network entity; buffer at least a portion of the media data in memory;compress the attribute map to form a compressed attribute map; and send the compressed attribute map and additional data corresponding to at least the buffered portion of the media data to the second network entity to enable the second network entity to perform the second part of the computation graph using the attribute map and the additional data.

[00116] Similarly, the second Petition 870250092796, dated 10 / 10 / 2025, page 62 / 207 55 / 87 device 280 represents an example of a device for processing attribute set data formed from media data, including: a memory configured to store attribute set data and media data; and a processing system comprising one or more processors implemented in a circuit array, wherein the processing system is configured to: receive a compressed attribute map including attributes extracted from media data and processed according to a first part of a computation graph and additional data from a second network entity, wherein the additional data corresponds to at least a portion of the media data that was buffered by the second network entity; decompress the compressed attribute map;and perform a second part of the computation graph on the attribute map using the additional data, where the second part of the computation graph corresponds to processing tasks to be performed by the first network entity.

[00117] Figure 11 is a call flow diagram that illustrates an example method that can be implemented by the system architecture of Figure 10 according to the techniques of this disclosure.

[00118] Initially, the first device 270 (e.g., a UE) can perform task network part 1 (290). That is, the first device 270 can extract a set of attributes from one or more media data images. Task network part 1 272 with the original split point can then process the input data (e.g., video frames) according to a first portion of a task network, to form a set of attributes. Petition 870250092796, dated 10 / 10 / 2025, page 63 / 207 56 / 87 intermediate. Part 1 of the task network with a new split point can also process the input data and generate a different intermediate attribute set. Resource compression unit 274 can compress the intermediate attribute set associated with the original split point and transmit the compressed intermediate attribute set to the second device 280 (292). The first device 270 can buffer both the attribute sets associated with the original split point for a certain number of frames in buffer 276 and the attribute sets associated with the second split point in buffer 278.

[00119] The attribute decompression unit 282 of the second device 280 can decode the intermediate attribute set. Then, task network part 284 can perform task network part 2 (294). Based on the analysis of the task results, the second device 280 can request the differently processed buffered attribute sets for the frames of the first device 270 from buffer 278 (296). For example, the second device 280 can detect a particular object from the attribute set. In response, the first device 270 can send the differently processed attribute sets for one or more images or video data frames from buffer 278 to the second device 280, according to the request (298).

[00120] The second device 280 can use the additional data (the attribute sets processed differently for one or more images or video data frames from buffer 278, in this example) to perform Petition 870250092796, dated 10 / 10 / 2025, page 64 / 207 57 / 87 additional inference processing (300). In this example, the second device 280 also requests that the first device 270 adjust the buffer size 276 and / or the buffer size 278 (302), for example, to increase or decrease the buffer size(s) based on the results of the additional inference processing. In response, the first device 270 may adjust the buffer size 276 and / or the buffer size 278 according to the request of the second device 280 (304).

[00121] In this way, the method in Figure 11 represents an example of a data processing method for a set of attributes formed from media data that includes: performing, by a first network entity (for example, first device 270 in Figure 10), a first part of a computation graph on a set of media data to form an attribute map, where the first part of the computation graph represents first processing tasks performed by the first network entity, and a second part of the computation graph corresponds to second processing tasks to be performed by a second network entity; buffering, by the first network entity, at least a portion of the media data; compressing, by the first network entity, the attribute map to form a compressed attribute map;and send, through the first entity, the compressed attribute map and additional data corresponding at least to the buffered portion of the media data to the second network entity to enable the second network entity to perform the second part of the computation graph using the attribute map and the additional data. Petition 870250092796, dated 10 / 10 / 2025, page 65 / 207 58 / 87

[00122] The method in Figure 11 also represents an example of a data processing method for a set of attributes formed from media data that includes: receiving, by a first network entity (e.g., second device 280 in Figure 10), a compressed attribute map including attributes extracted from media data and processed according to a first part of a computation graph and additional data from a second network entity, wherein the additional data corresponds at least to the portion of the media data that was buffered by the second network entity; decompressing, by the first network entity, the compressed attribute map; and performing, by the first network entity, a second part of the computation graph on the attribute map using the additional data, wherein the second part of the computation graph corresponds to processing tasks to be performed by the first network entity.

[00123] Several examples of the techniques of this disclosure are summarized in the following clauses:

[00124] Clause 1: A method for processing attribute set data formed from media data, wherein the method comprises: performing, by a first network entity, a first set of processing tasks from a series of processing tasks to be performed on a media dataset to form an attribute map, wherein the first set of processing tasks corresponds to tasks to be performed by the first network entity, wherein a second set of processing tasks from the series of processing tasks Petition 870250092796, dated 10 / 10 / 2025, p. 66 / 207 59 / 87 must be performed by a second network entity; buffer at least a portion of the media data by the first network entity; encode the attribute map by the first network entity to form an encoded attribute map; send the encoded attribute map and additional data corresponding to at least the buffered portion of the media data by the first entity to the second network entity to enable the second network entity to perform the second set of processing tasks using the attribute map and the additional data.

[00125] Clause 2: The method of clause 1, wherein the first network entity comprises a user equipment (UE), and the second network entity comprises an application server (AS).

[00126] Clause 3: The method of either clause 1 and 2, wherein the additional data comprises at least one buffered image or video frame, a set of attributes compressed to a lower degree than the encoded attribute map, enhancement data of the attributes of the encoded attribute map, or compressed attributes corresponding to at least the buffered portion of the media data at a different split point.

[00127] Clause 4: The method of any of clauses 1 to 3, wherein the sending of additional data comprises sending the additional data in response to a request for additional data from the second network entity.

[00128] Clause 5: The method of clause 4, where the request identifies images or video frames by at least one sequence number or a range of Petition 870250092796, dated 10 / 10 / 2025, p. 67 / 207 60 / 87 time.

[00129] Clause 6: The method of either clause 4 or 5, where the request specifies a target compression rate or bit rate for the additional data.

[00130] Clause 7: The method of any of clauses 4 to 6, wherein the request includes data that specifies at least one of a cause for the request or a priority for the request.

[00131] Clause 8: The method of any of clauses 1 to 7 which additionally comprises receiving a request to increase or decrease the size of a buffer used to buffer at least the media data portion.

[00132] Clause 9: The method of clause 8, where the request specifies the size according to a number of images or video frames or a number of attributes.

[00133] Clause 10: The method of either clause 8 or 9, where the request specifies the size according to a time period to be covered at least by the buffered portion of the media data.

[00134] Clause 11: The method of any of clauses 8 to 10 which additionally comprises increasing or decreasing at least the buffered portion of the media data according to the request.

[00135] Clause 12: The method of clause 1, wherein the first network entity comprises a user equipment (UE), and the second network entity comprises an application server (AS).

[00136] Clause 13: The method of clause 1, in Petition 870250092796, dated 10 / 10 / 2025, page 68 / 207 61 / 87 that the additional data comprises at least one buffered image or video frame, a set of attributes compressed to a lower degree than the encoded attribute map, enhancement data of the attributes of the encoded attribute map, or compressed attributes corresponding to at least the buffered portion of the media data at a different split point.

[00137] Clause 14: The method of clause 1, where the sending of additional data comprises sending the additional data in response to a request for additional data from the second network entity.

[00138] Clause 15: The method of clause 14, where the request identifies images or video frames by at least one sequence number or a time interval.

[00139] Clause 16: The method of clause 14, where the request specifies a target compression rate or bit rate for the additional data.

[00140] Clause 17: The method of clause 14, where the request includes data that specifies at least one of a cause for the request or a priority for the request.

[00141] Clause 18: The method of clause 1 which further comprises receiving a request to increase or decrease the size of a buffer used to buffer at least the media data portion.

[00142] Clause 19: The method of clause 18, where the request specifies the size according to a number of images or video frames or a number of attributes.

[00143] Clause 20: The method of clause 18, in Petition 870250092796, dated 10 / 10 / 2025, page 69 / 207 62 / 87 that the request specifies the size according to a time period to be covered at least by the buffered portion of the media data.

[00144] Clause 21: The method of clause 18 which additionally comprises increasing or decreasing at least the buffered portion of the media data according to the request.

[00145] Clause 22: A method for processing attribute set data formed from media data, wherein the method comprises: receiving, by a first network entity, a map of encoded attributes including attributes extracted from media data and additional data from a second network entity, wherein the additional data corresponds at least to the portion of the media data that was buffered by the second network entity; decoding, by the first network entity, the map of encoded attributes; and performing, by the first network entity, one or more processing tasks on the map of encoded attributes using the additional data.

[00146] Clause 23: The method of clause 22, wherein the first network entity comprises an application server (AS), and the second network entity comprises a user device (UE).

[00147] Clause 24: The method of either clause 22 or 23, wherein the additional data comprises at least one buffered image or video frame, a set of attributes compressed to a lower degree than the encoded attribute map, enhancement data of the attributes of the encoded attribute map, or compressed attributes corresponding to at least the portion Petition 870250092796, dated 10 / 10 / 2025, p. 70 / 207 63 / 87 stored in a buffer of media data at a different split point.

[00148] Clause 25: The method of any of clauses 22 to 24 which additionally comprises sending a request for the additional data to the second network entity.

[00149] Clause 26: The method of clause 25 which additionally comprises: specifying, in the request, identifiers for images or video frames by at least one sequence number or a time interval.

[00150] Clause 27: The method of either clause 25 and 26 which additionally comprises specifying, in the request, a target compression rate or bit rate for the additional data.

[00151] Clause 28: The method of any of clauses 25 to 27, whereby sending the request comprises, after processing the encoded attribute map alone, the determination that additional data is required.

[00152] Clause 29: The method of clause 28, wherein processing of the encoded attribute map alone includes at least one of detecting an object using the encoded attribute map or determining that a confidence value for the encoded attribute map is below a threshold.

[00153] Clause 30: The method of any of clauses 25 to 29, wherein the sending of the request comprises sending the request based on at least one of an estimated communication network quality, a predicted communication network quality, a computing capacity of the second network entity, or a latency requirement of one or more processing tasks. Petition 870250092796, dated 10 / 10 / 2025, page 71 / 207 64 / 87

[00154] Clause 31: The method of any of clauses 22 to 30 which additionally comprises sending a request to increase or decrease the size of a buffer used to buffer at least the media data portion.

[00155] Clause 32: The method of clause 31 which additionally comprises specifying, in the request, the size according to a series of images or video frames or various attributes.

[00156] Clause 33: The method of either clause 31 and 32 which additionally comprises specifying, in the request, the size according to a period of time to be covered at least by the buffered portion of the media data.

[00157] Clause 34: The method of clause 22, wherein the first network entity comprises an application server (AS), and the second network entity comprises a user device (UE).

[00158] Clause 35: The method of clause 22, wherein the additional data comprise at least one of the buffered images or video frames, a set of attributes compressed to a lower degree than the encoded attribute map, enhancement data of the attributes of the encoded attribute map, or compressed attributes corresponding to at least the buffered portion of the media data at a different split point.

[00159] Clause 36: The method of clause 22 which additionally comprises sending a request for the additional data to the second network entity.

[00160] Clause 37: The method of clause 36 that Petition 870250092796, dated 10 / 10 / 2025, page 72 / 207 65 / 87 additionally includes: specifying, in the request, identifiers for images or video frames by at least one sequence number or time interval.

[00161] Clause 38: The method of clause 36 which additionally comprises specifying, in the request, a target compression rate or bit rate for the additional data.

[00162] Clause 39: The method of clause 25, whereby sending the request comprises, after processing the map of coded attributes alone, the determination that additional data is required.

[00163] Clause 40: The method of clause 39, wherein processing of the encoded attribute map alone includes at least one of detecting an object using the encoded attribute map or determining that a confidence value for the encoded attribute map is below a threshold.

[00164] Clause 41: The method of clause 25, wherein the sending of the request comprises sending the request based on at least one of an estimated communication network quality, a predicted communication network quality, a computing capacity of the second network entity, or a latency requirement of one or more processing tasks.

[00165] Clause 42: The method of clause 22 which further comprises sending a request to increase or decrease the size of a buffer used to buffer at least the media data portion.

[00166] Clause 43: The method of clause 42 which additionally comprises specifying, in the request, the size according to a series of images or frames of Petition 870250092796, dated 10 / 10 / 2025, p. 73 / 207 66 / 87 video or multiple attributes.

[00167] Clause 44: The method of clause 42 which additionally comprises specifying, in the request, the size according to a period of time to be covered at least by the buffered portion of the media data.

[00168] Clause 45: A device for retrieving data from media, wherein the device comprises one or more means for carrying out the method of any of clauses 1 to 44.

[00169] Clause 46: The device of clause 45, wherein the one or more means comprise one or more processors implemented in a circuit assembly.

[00170] Clause 47: The apparatus of clause 45, wherein the apparatus comprises at least one of: an integrated circuit; a microprocessor; and a wireless communication device.

[00171] Clause 48: A computer-readable storage medium that has stored instructions which, when executed, cause one or more processors to perform the method of any of clauses 1 to 44.

[00172] Clause 49: A method for processing attribute set data formed from media data, wherein the method comprises: performing, by a first network entity, a first part of a computation graph on a media dataset to form an attribute map, wherein the first part of the computation graph represents first processing tasks performed by the first network entity, wherein a second part of the computation graph corresponds to second processing tasks to be performed by Petition 870250092796, dated 10 / 10 / 2025, page 74 / 207 67 / 87 a second network entity; buffer, by the first network entity, at least a portion of the media data; compress, by the first network entity, the attribute map to form a compressed attribute map; and send, by the first entity, the compressed attribute map and additional data corresponding to at least the buffered portion of the media data to the second network entity to enable the second network entity to perform the second part of the computation graph using the attribute map and the additional data.

[00173] Clause 50: The method of clause 49, wherein the first network entity comprises a user equipment (UE), and the second network entity comprises an application server (AS).

[00174] Clause 51: The method of clause 49, wherein the additional data comprise at least one buffered image or video frame, a set of compressed attributes to a lower degree than the compressed attribute map, enhancement data of the compressed attribute map attributes, or compressed attributes corresponding to at least the buffered portion of the media data at a different split point.

[00175] Clause 52: The method of clause 49, wherein the sending of the compressed attribute map comprises sending the compressed attribute map in a first instance, wherein the method further comprises receiving a request for additional data in a second instance subsequent to the first instance, wherein the sending of additional data comprises sending the additional data in response to the request for additional data in a third instance subsequent to the second instance. Petition 870250092796, dated 10 / 10 / 2025, p. 75 / 207 68 / 87

[00176] Clause 53: The method of clause 52, where the request identifies images or video frames by at least one sequence number or a time interval.

[00177] Clause 54: The method of clause 52, where the request specifies a target compression rate or bit rate for the additional data.

[00178] Clause 55: The method of clause 52, where the request includes data that specifies at least one of a cause for the request or a priority for the request.

[00179] Clause 56: The method of clause 49 which further comprises receiving a request to increase or decrease the size of a buffer used to buffer at least the media data portion.

[00180] Clause 57: The method of clause 56, where the request specifies the size according to a number of images or video frames or a number of attributes.

[00181] Clause 58: The method of clause 56, where the request specifies the size according to a time period to be covered at least by the buffered portion of the media data.

[00182] Clause 59: The method of clause 56 which additionally comprises increasing or decreasing at least the buffered portion of the media data according to the request.

[00183] Clause 60: The method of clause 49, where the first and second processing tasks include one or more of object detection, object tracking, or instance segmentation. Petition 870250092796, dated 10 / 10 / 2025, pp. 76 / 207 69 / 87

[00184] Clause 61: A device for processing attribute set data formed from media data, wherein the device comprises: a memory configured to store attribute set data and media data; and a processing system comprising one or more processors implemented in a circuit array, wherein the processing system is configured to: perform a first part of a computation graph on a set of media data to form an attribute map, wherein the first part of the computation graph represents first processing tasks performed by the device, wherein the device comprises a first network entity, wherein a second part of the computation graph corresponds to second processing tasks to be performed by a second network entity; buffer at least a portion of the media data in memory;compress the attribute map to form a compressed attribute map; and send the compressed attribute map and additional data corresponding to at least the buffered portion of the media data to the second network entity to enable the second network entity to perform the second part of the computation graph using the attribute map and the additional data.

[00185] Clause 62: The device of clause 61, wherein the device comprises a user equipment (UE), and the second network entity comprises an application server (AS).

[00186] Clause 63: The device of clause 61, in which the additional data comprises at least one of the buffered images or video frames, Petition 870250092796, dated 10 / 10 / 2025, page 77 / 207 70 / 87 a set of attributes compressed to a lower degree than the compressed attribute map, enhancement data of the compressed attribute map attributes, or compressed attributes corresponding at least to the buffered portion of the media data at a different split point.

[00187] Clause 64: The device of clause 61, wherein the processing system is configured to send the compressed attribute map at a first time, wherein the processing system is further configured to receive a request for additional data at a second time after the first time, and wherein the processing system is configured to send the additional data in response to the request for additional data at a third time after the second time.

[00188] Clause 65: The device of clause 64, wherein the request includes one or more of the following identifying images or video frames of data by at least one of sequence numbers or a time interval, data specifying a target compression rate or bit rate for the additional data, or data specifying at least one of a cause for the request or a priority for the request.

[00189] Clause 66: The device of clause 61, wherein the processing system is additionally configured to receive a request to increase or decrease the size of a buffer used to buffer at least the media data portion.

[00190] Clause 67: The clause's provision Petition 870250092796, dated 10 / 10 / 2025, pp. 78 / 207 71 / 87 66, where the request specifies the size according to a number of images or video frames or a number of attributes.

[00191] Clause 68: The device of clause 66, where the request specifies the size according to a time period to be covered at least by the buffered portion of the media data.

[00192] Clause 69: The device of clause 66, whereby the processing system is additionally configured to increase or decrease at least the buffered portion of the media data according to the request.

[00193] Clause 70: A method for processing attribute set data formed from media data, wherein the method comprises: receiving, by a first network entity, a compressed attribute map including attributes extracted from media data and processed according to a first part of a computation graph and additional data from a second network entity, wherein the additional data corresponds at least to the portion of the media data that was buffered by the second network entity; decompressing, by the first network entity, the compressed attribute map; and performing, by the first network entity, a second part of the computation graph on the attribute map using the additional data, wherein the second part of the computation graph corresponds to processing tasks to be performed by the first network entity.

[00194] Clause 71: The method of clause 70, wherein the first network entity comprises a server Petition 870250092796, dated 10 / 10 / 2025, pp. 79 / 207 72 / 87 of applications (AS), and the second network entity comprises a user equipment (UE).

[00195] Clause 72: The method of clause 70, wherein the additional data comprise at least one of the buffered images or video frames, a set of compressed attributes to a lower degree than the compressed attribute map, enhancement data of the compressed attribute map attributes, or compressed attributes corresponding to at least the buffered portion of the media data at a different split point.

[00196] Clause 73: The method of clause 70 which further comprises sending a request for additional data to the second network entity, wherein receiving the compressed attribute map comprises receiving the compressed attribute map at a first time, wherein sending the request comprises sending the request at a second time subsequent to the first time, and wherein receiving the additional data comprises receiving the additional data at a third time subsequent to the second time.

[00197] Clause 74: The method of clause 73 which additionally comprises: specifying, in the request, identifiers for images or video frames by at least one sequence number or a time interval.

[00198] Clause 75: The method of clause 73 which additionally comprises specifying, in the request, a target compression rate or bit rate for the additional data.

[00199] Clause 76: The method of clause 73, in which the sending of the request comprises, after processing the compacted attribute map alone, the determination of Petition 870250092796, dated 10 / 10 / 2025, pp. 80 / 207 73 / 87 that additional data is required.

[00200] Clause 77: The method of clause 76, wherein the processing of the compressed attribute map alone includes at least one of detecting an object using the compressed attribute map or determining that a confidence value for the compressed attribute map is below a threshold.

[00201] Clause 78: The method of clause 73, wherein the sending of the request comprises sending the request based on at least one of an estimated communication network quality, a predicted communication network quality, a computing capacity of the second network entity, or a latency requirement of one or more processing tasks.

[00202] Clause 79: The method of clause 70 which further comprises sending a request to increase or decrease the size of a buffer used to buffer at least the media data portion.

[00203] Clause 80: The method of clause 79 which additionally comprises specifying, in the request, the size according to a series of images or video frames or various attributes.

[00204] Clause 81: The method of clause 79 which additionally comprises specifying, in the request, the size according to a period of time to be covered at least by the buffered portion of the media data.

[00205] Clause 82: The method of clause 70, wherein the processing tasks include at least one of object detection, object tracking, or instance segmentation.

[00206] Clause 83: A device for Petition 870250092796, dated 10 / 10 / 2025, page 81 / 207 74 / 87 processing attribute set data formed from media data, wherein the device comprises: a memory configured to store attribute set data and media data; and a processing system comprising one or more processors implemented in a circuit array, wherein the processing system is configured to: receive a compressed attribute map including attributes extracted from media data and processed according to a first part of a computation graph and additional data from a second network entity, wherein the additional data corresponds to at least a portion of the media data that was buffered by the second network entity; decompress the compressed attribute map;and perform a second part of the computation graph on the attribute map using the additional data, where the second part of the computation graph corresponds to processing tasks to be performed by the first network entity.

[00207] Clause 84: The device of clause 83, in which the processing system is configured to receive the compressed attribute map at a first time, send a request for additional data at a second time after the first time, and receive the additional data at a third time after the second time.

[00208] Clause 85: A method for processing attribute set data formed from media data, wherein the method comprises: performing, by a first network entity, a first part of a computation graph on a media dataset to form Petition 870250092796, dated 10 / 10 / 2025, page 82 / 207 75 / 87 an attribute map, wherein the first part of the computation graph represents initial processing tasks performed by the first network entity, wherein a second part of the computation graph corresponds to secondary processing tasks to be performed by a second network entity; buffering, by the first network entity, at least a portion of the media data; compressing, by the first network entity, the attribute map to form a compressed attribute map; and sending, by the first entity, the compressed attribute map and additional data corresponding to at least the buffered portion of the media data to the second network entity to enable the second network entity to perform the second part of the computation graph using the attribute map and the additional data.

[00209] Clause 86: The method of clause 85, wherein the first network entity comprises a user equipment (UE), and the second network entity comprises an application server (AS).

[00210] Clause 87: The method of either clause 85 or 86, wherein the additional data comprises at least one buffered image or video frame, a set of compressed attributes to a lower degree than the compressed attribute map, enhancement data of the compressed attribute map attributes, or compressed attributes corresponding to at least the buffered portion of the media data at a different split point.

[00211] Clause 88: The method of any of clauses 85 to 87, where the attribute map is sent. Petition 870250092796, dated 10 / 10 / 2025, p. 83 / 207 76 / 87 compressed involves sending the compressed attribute map in a first instance, wherein the method additionally involves receiving a request for additional data in a second instance subsequent to the first instance, wherein the sending of additional data involves sending the additional data in response to the request for additional data in a third instance subsequent to the second instance.

[00212] Clause 89: The method of clause 88, where the request identifies images or video frames by at least one sequence number or time interval.

[00213] Clause 90: The method of either clause 88 or 89, where the request specifies a target compression rate or bit rate for the additional data.

[00214] Clause 91: The method of any of clauses 88 to 90, wherein the request includes data that specifies at least one of a cause for the request or a priority for the request.

[00215] Clause 92: The method of any of clauses 85 to 91 which additionally comprises receiving a request to increase or decrease the size of a buffer used to buffer at least the media data portion.

[00216] Clause 93: The method of clause 92, where the request specifies the size according to a number of images or video frames or a number of attributes.

[00217] Clause 94: The method of either clause 92 or 93, where the request specifies the size according to a time period to be covered at least by the buffered portion of the media data. Petition 870250092796, dated 10 / 10 / 2025, p. 84 / 207 77 / 87

[00218] Clause 95: The method of any of clauses 92 to 94 which additionally comprises increasing or decreasing at least the buffered portion of the media data according to the request.

[00219] Clause 96: The method of any of clauses 85 to 95, wherein the first and second processing tasks include one or more of object detection, object tracking, or instance segmentation.

[00220] Clause 97: A device for processing attribute set data formed from media data, wherein the device comprises: a memory configured to store attribute set data and media data; and a processing system comprising one or more processors implemented in a circuit array, wherein the processing system is configured to: perform a first part of a computation graph on a set of media data to form an attribute map, wherein the first part of the computation graph represents first processing tasks performed by the device, wherein the device comprises a first network entity, wherein a second part of the computation graph corresponds to second processing tasks to be performed by a second network entity; buffer at least a portion of the media data in memory;compress the attribute map to form a compressed attribute map; and send the compressed attribute map and additional data corresponding at least to the buffered portion of the media data to the second network entity to enable the second network entity to perform the second part of the computation graph; Petition 870250092796, dated 10 / 10 / 2025, page 85 / 207 78 / 87 using the attribute map and additional data.

[00221] Clause 98: The device of clause 97, wherein the processing system is configured to send the compressed attribute map at a first time, wherein the processing system is further configured to receive a request for additional data at a second time after the first time, and wherein the processing system is configured to send the additional data in response to the request for additional data at a third time after the second time.

[00222] Clause 99: The device of either clause 97 and 98, wherein the additional data comprises at least one buffered image or video frame, a set of compressed attributes to a lower degree than the compressed attribute map, enhancement data of the compressed attribute map attributes, or compressed attributes corresponding to at least the buffered portion of the media data at a different split point.

[00223] Clause 100: A method for processing attribute set data formed from media data, wherein the method comprises: receiving, by a first network entity, a compressed attribute map including attributes extracted from media data and processed according to a first part of a computation graph and additional data from a second network entity, wherein the additional data corresponds at least to the portion of the media data that was buffered by the second network entity; Petition 870250092796, dated 10 / 10 / 2025, page 86 / 207 79 / 87 to decompress, by the first network entity, the compressed attribute map; and to perform, by the first network entity, a second part of the computation graph on the attribute map using the additional data, where the second part of the computation graph corresponds to processing tasks to be performed by the first network entity.

[00224] Clause 101: The method of clause 100, wherein the first network entity comprises an application server (AS), and the second network entity comprises a user device (UE).

[00225] Clause 102: The method of either clause 100 or 101, wherein the additional data comprises at least one buffered image or video frame, a set of compressed attributes to a lower degree than the compressed attribute map, enhancement data of the compressed attribute map attributes, or compressed attributes corresponding to at least the buffered portion of the media data at a different split point.

[00226] Clause 103: The method of any of clauses 100 to 102 which additionally comprises sending a request for additional data to the second network entity, wherein receiving the compressed attribute map comprises receiving the compressed attribute map at a first time, wherein sending the request comprises sending the request at a second time subsequent to the first time, and wherein receiving the additional data comprises receiving the additional data at a third time subsequent to the second time. Petition 870250092796, dated 10 / 10 / 2025, page 87 / 207 80 / 87

[00227] Clause 104: The method of clause 103 which additionally comprises: specifying, in the request, identifiers for images or video frames by at least one sequence number or a time interval.

[00228] Clause 105: The method of either clause 103 and 104 comprising additionally specifying, in the request, a target compression rate or bit rate for the additional data.

[00229] Clause 106: The method of any of clauses 103 to 105, wherein the sending of the request comprises, after processing the compressed attribute map alone, the determination that additional data is required.

[00230] Clause 107: The method of clause 106, wherein the processing of the compressed attribute map alone includes at least one of detecting an object using the compressed attribute map or determining that a confidence value for the compressed attribute map is below a threshold.

[00231] Clause 108: The method of any of clauses 100 to 107, wherein the sending of the request comprises sending the request based on at least one of an estimated communication network quality, a predicted communication network quality, a computing capacity of the second network entity, or a latency requirement of one or more processing tasks.

[00232] Clause 109: The method of any of clauses 100 to 108 which additionally comprises sending a request to increase or decrease the size of a buffer used to buffer at least the media data portion. Petition 870250092796, dated 10 / 10 / 2025, pp. 88 / 207 81 / 87

[00233] Clause 110: The method of clause 109 which additionally comprises specifying, in the request, the size according to a series of images or video frames or various attributes.

[00234] Clause 111: The method of either clause 109 and 110 which additionally comprises specifying, in the request, the size according to a period of time to be covered at least by the buffered portion of the media data.

[00235] Clause 112: The method of any of clauses 100 to 111, wherein the processing tasks include at least one of object detection, object tracking, or instance segmentation.

[00236] Clause 113: A device for processing attribute set data formed from media data, wherein the device comprises: a memory configured to store attribute set data and media data; and a processing system comprising one or more processors implemented in a circuit array, wherein the processing system is configured to: receive a compressed attribute map including attributes extracted from media data and processed according to a first part of a computation graph and additional data from a second network entity, wherein the additional data corresponds to at least a portion of the media data that has been buffered by the second network entity; decompress the compressed attribute map; and perform a second part of the computation graph on the attribute map using the additional data, wherein the second part of the computation graph Petition 870250092796, dated 10 / 10 / 2025, pp. 89 / 207 82 / 87 corresponds to processing tasks to be performed by the first network entity.

[00237] Clause 114: The device of clause 113, in which the processing system is configured to receive the compressed attribute map at a first time, send a request for additional data at a second time after the first time, and receive the additional data at a third time after the second time.

[00238] Clause 115: A device for processing attribute set data formed from media data, wherein the device comprises: means for performing a first part of a computation graph on a media data set to form an attribute map, wherein the first part of the computation graph represents first processing tasks performed by the device, wherein a second part of the computation graph corresponds to second processing tasks to be performed by a second network entity separate from the device; means for buffering at least a portion of the media data; means for compressing the attribute map to form a compressed attribute map;and means to send the compressed attribute map and additional data corresponding at least to the buffered portion of the media data to the second network entity to enable the second network entity to perform the second part of the computation graph using the attribute map and the additional data.

[00239] Clause 116: A data processing device with a set of attributes formed from media data, wherein the device comprises: Petition 870250092796, dated 10 / 10 / 2025, pp. 90 / 207 83 / 87 means for receiving a compressed attribute map including attributes extracted from media data and processed according to a first part of a computation graph and additional data from a second network entity, wherein the additional data corresponds at least to the portion of the media data that was buffered by the second network entity; means for decompressing the compressed attribute map; and means for performing a second part of the computation graph on the attribute map using the additional data, wherein the second part of the computation graph corresponds to processing tasks to be performed by the first network entity.

[00240] Clause 117: A computer-readable storage medium that has stored therein instructions which, when executed, cause a processor of a first network entity to: perform a first part of a computation graph on a set of media data to form an attribute map, wherein the first part of the computation graph represents first processing tasks performed by the device, wherein the device comprises a first network entity, wherein a second part of the computation graph corresponds to second processing tasks to be performed by a second network entity; buffer at least a portion of the media data in memory; compress the attribute map to form a compressed attribute map;and send the compressed attribute map and additional data corresponding to at least the buffered portion of the media data to the second network entity to enable the second network entity to perform the second part of the computation graph; Petition 870250092796, dated 10 / 10 / 2025, pp. 91 / 207 84 / 87 using the attribute map and additional data.

[00241] Clause 118: A computer-readable storage medium that has stored therein instructions which, when executed, cause a processor of a first network entity to: receive, by a first network entity, a compressed attribute map including attributes extracted from media data and processed according to a first part of a computation graph and additional data from a second network entity, wherein the additional data corresponds at least to the portion of the media data that was buffered by the second network entity; decompress the compressed attribute map; and perform a second part of the computation graph on the attribute map using the additional data, wherein the second part of the computation graph corresponds to processing tasks to be performed by the first network entity.

[00242] In one or more examples, the functions described may be implemented in hardware, software, firmware, or any combination thereof. If implemented in software, the functions may be stored or transmitted as one or more instructions or code in a computer-readable medium and executed by a hardware-based processing unit. Computer-readable media may include computer-readable storage media, which correspond to a tangible medium, such as data storage media, or communication media, including any medium that facilitates the transfer of a computer program from one place to another, for example, according to a communication protocol. In this way, computer-readable media may generally correspond to (1) media Petition 870250092796, dated 10 / 10 / 2025, page 92 / 207 85 / 87 tangible computer-readable storage media that are non-transient or (2) a communication medium, such as a signal or a carrier wave. Data storage media may be any available media that can be accessed by one or more computers or one or more processors to retrieve instructions, code, and / or data structures for the implementation of the techniques described in this disclosure. A computer program product may include a computer-readable medium.

[00243] By way of example, and not limitation, such computer-readable storage media may comprise random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), CD-ROM or other optical disk storage, magnetic disk storage or other magnetic storage devices, flash memory or any other media that can be used to store the desired program code in the form of instructions or data structures and that can be accessed by a computer. In addition, any connection is properly termed a computer-readable medium.For example, if instructions are transmitted from a website, server, or other remote source using a coaxial cable, fiber optic cable, twisted pair cable, digital subscriber line (DSL), or wireless technologies such as infrared, radio, and microwave, then the coaxial cable, fiber optic cable, twisted pair cable, DSL, or wireless technologies such as infrared, radio, and microwave... Petition 870250092796, dated 10 / 10 / 2025, pp. 93 / 207 86 / 87 waves, will be included in the definition of media. It should be understood, however, that computer-readable storage media and data storage media do not include connections, carrier waves, signals, or other transient media, but instead refer to tangible non-transient storage media. As used in the present invention, disks (disk and disc) include compact disc (CD), laser disc, optical disc, digital versatile disc (DVD), floppy disk, and Blu-ray disc, where disks generally reproduce data magnetically, while discs reproduce data optically by means of lasers. Combinations of the above items should also be included in the scope of computer-readable media.

[00244] The instructions may be executed by one or more processors, such as one or more digital signal processors (DSPs), general-purpose microprocessors, application-specific integrated circuits (ASICs), field-programmable logic arrays (FPGAs), or other equivalent set of discrete or integrated logic circuits. Consequently, the term processor, as used in the present invention, may refer to any of the aforementioned structures or any other structure suitable for implementing the techniques described in the present invention. Furthermore, in some respects, the functionality described in the present invention may be provided in dedicated hardware and / or software modules configured for encoding and decoding, or may be incorporated into a combined codec. Additionally, the techniques may be fully implemented in one or more logic circuits or elements. Petition 870250092796, dated 10 / 10 / 2025, pp. 94 / 207 87 / 87

[00245] The techniques in this disclosure can be implemented in a wide variety of devices or appliances, including a wireless handset, an integrated circuit (IC), or a set of ICs (e.g., a chipset). Several components, modules, or units are described in this disclosure to emphasize functional aspects of devices configured to perform the disclosed techniques, but do not necessarily require implementation by different hardware units. Instead, as described above, several units can be combined into a codec hardware unit or provided by a collection of interoperable hardware units, including one or more processors, as described above, in conjunction with suitable software and / or firmware.

[00246] Several examples have been described. These and other examples are within the scope of the following claims. Petition 870250092796, dated 10 / 10 / 2025, pp. 95 / 207

Claims

1 / 8 CLAIMS 1. A method for processing data from a set of attributes formed from media data, the method being characterized by comprising: performing, by a first network entity, a first part of a computation graph on a set of media data to form an attribute map, wherein the first part of the computation graph represents first processing tasks performed by the first network entity, wherein a second part of the computation graph corresponds to second processing tasks to be performed by a second network entity; buffering, by the first network entity, at least a portion of the media data; compressing, by the first network entity, the attribute map to form a compressed attribute map;and send, through the first entity, the compressed attribute map and additional data corresponding at least to the buffered portion of the media data to the second network entity to enable the second network entity to perform the second part of the computation graph using the attribute map and the additional data.

2. A method according to claim 1, characterized in that the first network entity comprises a user equipment (UE), and the second network entity comprises an application server (AS).

3. Method, according to claim 1, characterized in that the additional data comprises at least one buffered image or video frame, a set of compressed attributes to a lower degree than the compressed attribute map, enhancement data of the compressed attribute map attributes, or compressed attributes corresponding to at least the buffered portion of the media data at a different split point.

4. A method according to claim 1, characterized in that sending the compressed attribute map comprises sending the compressed attribute map at a first time, wherein the method further comprises receiving a request for additional data at a second time subsequent to the first time, wherein sending the additional data comprises sending the additional data in response to the request for additional data at a third time subsequent to the second time.

5. Method, according to claim 4, characterized in that the application identifies images or video frames by at least one sequence number or a time interval.

6. Method according to claim 4, characterized in that the application specifies a target compression rate or bit rate for the additional data.

7. Method, according to claim 4, characterized in that the request includes data that specifies at least one of a cause for the request or a priority for the request.

8. A method according to claim 1, characterized by further comprising receiving a request to increase or decrease the size of a buffer used to buffer at least the media data portion.

9. Method according to claim 8, Petition 870250092796, dated 10 / 10 / 2025, pp. 185 / 207 3 / 8 characterized in that the application specifies the size according to a number of images or video frames or a number of attributes.

10. Method, according to claim 8, characterized in that the request specifies the size according to a time period to be covered at least by the buffered portion of the media data.

11. Method according to claim 8, characterized by further comprising increasing or decreasing at least the buffered portion of the media data as requested.

12. A method according to claim 1, characterized in that the first and second processing tasks include one or more of object detection, object tracking, or instance segmentation.

13. Method, according to claim 1, characterized in that at least the portion of the media data encompasses a round-trip travel time between the first network entity and the second network entity and a time for the second network entity to perform the second part of the computation graph.

14. Device for processing attribute set data formed from media data, the device being characterized by comprising: a memory configured to store attribute set data and media data; and a processing system comprising one or more processors implemented in a circuit assembly, wherein the processing system is configured to: perform a first part of a graph of Petition 870250092796, of 10 / 10 / 2025, page.186 / 207 4 / 8 computation on a set of media data to form an attribute map, wherein the first part of the computation graph represents first processing tasks performed by the device, wherein the device comprises a first network entity, wherein a second part of the computation graph corresponds to second processing tasks to be performed by a second network entity; buffer at least a portion of the media data in memory; compress the attribute map to form a compressed attribute map; and send the compressed attribute map and additional data corresponding at least to the buffered portion of the media data to the second network entity to enable the second network entity to perform the second part of the computation graph using the attribute map and the additional data.

15. Device according to claim 14, characterized in that the processing system is configured to send the compressed attribute map at a first time, wherein the processing system is further configured to receive a request for additional data at a second time subsequent to the first time, and wherein the processing system is configured to send the additional data in response to the request for additional data at a third time subsequent to the second time.

16. Device according to claim 15, characterized in that the application includes one or more of the following data identification images or video frames, as per Petition 870250092796, dated 10 / 10 / 2025, page 187 / 207 5 / 8, at least one of the following sequence numbers or a time interval, data specifying a target compression rate or bit rate for the additional data, or data specifying at least one of the following reasons for the application or a priority for the application.

17. A method for processing attribute set data formed from media data, the method being characterized by comprising: receiving, by a first network entity, a compressed attribute map including attributes extracted from media data and processed according to a first part of a computation graph and additional data from a second network entity, wherein the additional data corresponds to at least a portion of the media data that was buffered by the second network entity; decompressing, by the first network entity, the compressed attribute map; and performing, by the first network entity, a second part of the computation graph on the attribute map using the additional data, wherein the second part of the computation graph corresponds to processing tasks to be performed by the first network entity.

18. Method according to claim 17, characterized in that the first network entity comprises an application server (AS), and the second network entity comprises a user device (UE).

19. Method, according to claim 17, characterized in that the additional data comprises at least one buffered image or video frame, a set of compressed attributes to a lower degree than the compressed attribute map, enhancement data of the compressed attribute map attributes, or compressed attributes corresponding to at least the buffered portion of the media data at a different split point.

20. A method according to claim 17, characterized by further comprising sending a request for additional data to a second network entity, wherein receiving the compressed attribute map comprises receiving the compressed attribute map at a first time, wherein sending the request comprises sending the request at a second time subsequent to the first time, and wherein receiving the additional data comprises receiving the additional data at a third time subsequent to the second time.

21. Method according to claim 20, characterized by further comprising specifying, in the application, identifiers for images or video frames by at least one sequence number or time interval.

22. Method according to claim 20, characterized by further comprising specifying, in the application, a target compression rate or bit rate for the additional data.

23. Method, according to claim 20, characterized in that the sending of the request comprises, after processing the compressed attribute map alone, the determination that additional data is needed.

24. Method, according to claim 23, characterized in that the processing of the compressed attribute map alone includes at least one of detecting an object using the compressed attribute map or determining Petition 870250092796, dated 10 / 10 / 2025, pp. 189 / 207 7 / 8 that a confidence value for the compressed attribute map is below a threshold.

25. Method, according to claim 20, characterized in that the sending of the request comprises sending the request based on at least one of an estimated communication network quality, a predicted communication network quality, a computing capacity of the second network entity, or a latency requirement of one or more processing tasks.

26. A method according to claim 17, characterized by further comprising sending a request to increase or decrease the size of a buffer used to buffer at least the media data portion.

27. Method, according to claim 26, characterized by further comprising specifying, in the application, the size according to a series of images or video frames or various attributes.

28. Method according to claim 26, characterized by further comprising specifying, in the application, the size according to a time period to be covered at least by the buffered portion of the media data.

29. A method according to claim 17, characterized in that the processing tasks include at least one of object detection, object tracking, or instance segmentation.

30. Device for processing data from a set of attributes formed from media data, the device being characterized by comprising: Petition 870250092796, dated 10 / 10 / 2025, page.190 / 207 8 / 8 a memory configured to store attribute set data and media data; and a processing system comprising one or more processors implemented in a circuit set, wherein the processing system is configured to: receive a compressed attribute map including attributes extracted from media data and processed according to a first part of a computation graph and additional data from a second network entity, wherein the additional data corresponds to at least a portion of the media data that was buffered by the second network entity; decompress the compressed attribute map; and perform a second part of the computation graph on the attribute map using the additional data, wherein the second part of the computation graph corresponds to processing tasks to be performed by the first network entity. Petition 870250092796, dated 10 / 10 / 2025, pp. 191 / 207.

Images