43 results about "Dynamic resolution" patented technology

The invention discloses a VR rendering optimization method and system based on eyeball tracking, a medium and electronic equipment. The method comprises the following steps: preprocessing VR data to obtain target data; predicting a fixation center point at the next moment according to the target data and based on the trained bidirectional LSTM time sequence neural network; setting dynamic resolution mapping of VR rendering according to the fixation center point; setting a plurality of rendering areas for VR rendering according to the fixation center point; dynamically adjusting target dynamic quantization parameters respectively corresponding to the plurality of rendering areas based on the current network condition; vR dynamic rendering at the next moment is carried out based on the dynamic resolution mapping and the target dynamic quantization parameter corresponding to each rendering area; therefore, the rendering task is allocated in advance according to the predicted fixation center point, so that the user is prevented from feeling dizzy; and meanwhile, dynamic resolution mapping is adopted for rendering, so that the splitting feeling caused by traditional partitioning can be avoided while the rendering calculation amount is reduced.

The invention discloses a multi-scene micro garbage real-time detection method and a multi-scene micro garbage real-time detection system, and aims to solve the defects of an existing garbage detection method in the aspects of small target identification, environmental adaptability and system integration. According to the method, on the basis of an improved YOLOv8 architecture, a Transform module and an efficient channel attention mechanism are fused, and the tiny target feature extraction and multi-scale information fusion capability is enhanced. The system adopts dynamic resolution reasoning, INT8 quantitative compression and time-space domain post-processing, so that the detection precision and the reasoning efficiency are remarkably improved. Experiments prove that the system effectively improves the small target recall rate and environmental adaptability in a garbage detection task, comprises eight corresponding modules, realizes full-process automatic detection, has relatively high real-time performance and automation level, is suitable for various application scenes such as smart city environment monitoring, and has important practical application value.

The invention provides a road preview method and system based on a point cloud dynamic resolution allocation algorithm, and belongs to the field of automatic driving and intelligent traffic systems. The method comprises the following steps: acquiring an image pair acquired by a binocular camera, and preprocessing the image pair; generating a depth map of the scene based on the preprocessed image pair, and dividing the depth map into a plurality of regions of interest in a depth dimension; for each region of interest, calculating the importance weight of the region of interest based on the depth gradient and the motion vector; dynamically distributing the point cloud resolution of the region through a mapping function according to the depth value of the region of interest and the importance weight corresponding to the depth value; and based on the distributed point cloud resolution, generating point cloud data of dynamic resolution for road preview and path planning of the vehicle. According to the method, the problems of resource waste and poor real-time performance caused by full-scene uniform sampling of traditional binocular vision are solved, and the precision and efficiency of automatic driving road preview are remarkably improved.

A computing system for generating a gain map-based enhanced High Dynamic Resolution (HDR) image receives an HDR image comprising a plurality of pixels, each pixel having one or a plurality of brightness values for each of a plurality of color components, applies at least a first tone mapping algorithm to each pixel in the HDR image, generates a gain map using the HDR image and the SDR image, generates a gain map-based enhanced HDR image based on the SDR image and the gain map, and generates an output based on the gain map-based enhanced HDR image.

The invention relates to the technical field of artificial intelligence, and particularly provides a front-end code generation method and device, equipment, a storage medium and a program product.The front-end code generation method comprises the steps that a user interface image and a user instruction text are obtained; target visual features are generated based on the user interface image through a visual encoder, the target visual features comprise a first visual feature and a second visual feature, the first visual feature is a visual feature obtained by performing visual encoding processing based on a fixed resolution on the user interface image, and the second visual feature is a visual feature obtained by performing visual encoding processing based on the fixed resolution on the user interface image; the second visual feature is a visual feature obtained by performing dynamic resolution-based visual coding processing on the user interface image; and through a large language model, generating a front-end code based on the target visual features and the user instruction text. By implementing the technical scheme of the embodiment of the invention, the accuracy of front-end code generation can be improved.

The invention provides a 3D laser SLAM method and device based on optimized loopback detection and a medium, and relates to the technical field of simultaneous positioning and mapping of robots, and the method comprises the following steps: (1) a data preprocessing step: receiving original data of an IMU and a laser radar Lidar, and carrying out the denoising and time synchronization processing; (2) a feature extraction and map construction step: extracting angular point and surface point features from the preprocessed data, performing pre-integration on IMU data to obtain an initial pose, and performing registration on a current Lidar point cloud and a local map to construct a global map; (3) loopback detection: generating an enhanced descriptor based on the key frame point cloud, and constructing a KD-Tree structure with dynamic resolution for storage and indexing; and carrying out rough search on loopback candidate frames by inquiring the KD-Tree, carrying out two-stage pose matching on the candidate frames to confirm loopback, and dynamically distributing the weight of loopback constraint according to the environmental complexity. The loopback detection success rate can be improved, and dynamic scene mapping drift is reduced.

The invention discloses an edge end real-time target detection method based on multi-granularity attention fusion and space-time consistency constraint. The method aims at solving the problems that existing edge end target detection does not give consideration to real-time performance and precision, and video sequence detection consistency is poor. The method comprises the following steps that S1, a multi-source data set is constructed and preprocessed; s2, designing and optimizing a lightweight multi-granularity feature extraction backbone network; s3, implementing a dynamic multi-granularity attention fusion module, and performing adaptive fusion on multi-granularity features; s4, enhancing the video sequence detection consistency through a predictive spatio-temporal feature alignment module and a spatio-temporal consistency loss function; s5, constructing a detection head and a multi-task loss function; s6, performing model training; and S7, carrying out edge device adaptation and reasoning acceleration by adopting quantitative perception training, dynamic resolution adjustment, optimized non-maximum suppression and other methods. According to the invention, the precision, the time sequence consistency and the real-time performance of edge end real-time target detection are obviously improved, and the method has the characteristic of high efficiency.

A method for reducing artifacts in streaming video implemented by a computing device involves determining whether a video quality assessment score for a current segment of a channel of a video stream with a first resolution and a first bit rate meets a quality threshold, replacing the current segment with a replacement segment with a lower resolution than the first resolution, in response to the current segment failing to meet the quality threshold, and publishing the replacement segment at the first bit rate as part of the video stream.

The present disclosure provides a multimodal data fusion method, device and vehicle based on dynamic Gaussian modeling. In one embodiment of the present disclosure, the multimodal data fusion method based on dynamic Gaussian modeling includes: acquiring continuous multi-frame multimodal data, obtaining 3D voxel features and multi-view image features through the multimodal data, constructing a dynamic Gaussian parameterized model based on the 3D voxel features, generating multi-view Gaussian features using the dynamic Gaussian parameterized model and multi-view image features, and projecting the multi-view Gaussian features into a dynamic resolution BEV space to obtain BEV features, wherein the 3D voxel features are obtained by dynamic voxelization, and the dynamic voxelization is based on octree sparse voxel coding and adaptively adjusts the resolution according to the complexity of the scene. The present disclosure can optimize the utilization of computing resources while ensuring perception accuracy.

An example device for retrieving media data includes a memory configured to store video data; a video decoder configured to decode the video data; and one or more processors implemented in circuitry and configured to: determine that a media presentation includes first video data at a first spatial resolution and second video data at a second spatial resolution, the second spatial resolution being different than the first spatial resolution; receive a first portion of the first video data at the first spatial resolution for a first playback time; send the first portion of the first video data at the first spatial resolution to the video decoder; receive a second portion of the second video data at the second spatial resolution for a second playback time later than the first playback time; and send the second portion of the second video data at the second spatial resolution to the video decoder.

An apparatus, the apparatus comprising: a screen generator configured to generate a screen in response to an application; a resolution selector configured to select a resolution for rendering the screen based on a set of parameters associated with the screen; and a graphics processing unit (GPU) configured to render the screen on the display based on the selected resolution.

This application discloses a configuration method and electronic device for a server startup interface, relating to the field of server technology. The method includes initializing the baseboard management controller and allocating memory-mapped access resources for the video image array device to ensure the startup interface starts synchronously with hardware initialization; obtaining the horizontal and vertical axis percentage parameters of each display area, calculating the total area ratio of the display area, and, if the total area ratio meets a preset value, calculating the number of rows and columns of each display area based on the screen resolution to achieve dynamic resolution adaptation; after locating the position of each area using the memory-mapped access resources, controlling the video image array device to display the area precisely according to the percentage set by the user. This allows users to define which area displays what and what percentage it occupies, personalizing the startup interface content and layout to meet diverse user needs.

A mechanism for processing video data is disclosed. The mechanism includes determining to select one or more adaptive scaling filters. A conversion is performed between a visual media data and a bitstream based on the adaptive scaling filters.

The present disclosure relates to a static graph mode graph generation system supporting dynamic resolution. The system comprises: an original logic graph construction component, for a first specific graph generation task, tracking original computing logic by executing a construction function of a graph to construct an original logic graph; a logic graph optimization component, forming a first logic graph for the constructed original logic graph; a node insertion component, inserting a communication node and a runtime control node required by distributed computing for the first logic graph to generate a first runtime logic graph for multi-device collaboration; a runtime plan generation component, generating a first runtime execution plan based on a data dependency relationship of the first runtime logic graph; and a runtime initialization component, initializing a first specific graph runtime instance according to the first runtime execution plan to complete dynamic allocation of computing resources and configuration of an execution environment.

The application discloses a VR rendering optimization method and system based on eyeball tracking, a medium and an electronic device, and the method comprises the following steps: preprocessing VR data to obtain target data; predicting a gaze center point at the next moment according to the target data and based on a trained bidirectional LSTM time sequence neural network; setting a dynamic resolution mapping of VR rendering according to the gaze center point; setting a plurality of rendering areas of the VR rendering according to the gaze center point; dynamically adjusting target dynamic quantization parameters corresponding to the plurality of rendering areas respectively based on the current network condition; performing VR dynamic rendering at the next moment based on the dynamic resolution mapping and the target dynamic quantization parameters corresponding to each rendering area respectively; therefore, the application allocates rendering tasks in advance according to the predicted gaze center point, so that the user can avoid dizziness; meanwhile, the rendering adopts dynamic resolution mapping, so that the rendering calculation amount can be reduced, and the fragmentation caused by traditional partitioning can be avoided.

The invention relates to the field of image communication, in particular to a mobile terminal video stream data adaptation processing method based on an external camera, which comprises the following steps: an acquisition terminal ensures that a source frame rate is stable through the external camera with an IMU (Inertial Measurement Unit) in combination with frame rate monitoring and dynamic caching; the edge node adopts parallel preprocessing, a lightweight model and adaptive coding to compress single frame processing delay; the transmission link realizes fragmentation transmission, priority scheduling and multi-path redundancy based on UDP + RTSP, HTTP / 3 and SRT protocols; the terminal side reduces the decoding load through pre-decoding, fragmentation caching and dynamic resolution adaptation; and the full-link routing collaborative dispatching center monitors the indexes in real time and triggers a grading, degrading and fault switching mechanism. According to the technical scheme, high-stability and low-delay video stream processing with the end-to-end delay smaller than or equal to 150 ms, the frame rate fluctuation within + / -1 fps and the lagging rate smaller than or equal to 0.5% is achieved, and the immersive video experience in weak network and high load scenes is remarkably improved.

According to aspects of the disclosure, a method is provided for use in a sensor, the method comprising: detecting that a frequency of a first signal is in a first range, the first signal being generated, at least in part, by one or more first magnetic field sensing elements, the first signal being generated in response to a magnetic field that is associated with a rotating target, the rotating target including a plurality of pole pairs; identifying a first resolution that corresponds to the first range and causing the first resolution to become a current resolution of the sensor; and transmitting a data stream in accordance with the current resolution of the sensor, wherein transmitting the data stream includes: (i) transmitting a plurality of speed pulses that encode a speed of the rotating target, and (ii) transmitting a plurality of data pulse sets.

The application discloses a dynamic resolution ultrasonic phased array haptic generation device and a control method, relates to the technical field of ultrasonic haptic feedback, and realizes the software adjustment of the ultrasonic haptic focus size through an EDM beam forming algorithm, thereby breaking through the limitation of the traditional DAS method which is limited by hardware. Users can dynamically select a high resolution mode or a high radiation force mode according to specific application scenarios, and the flexibility and reality of haptic interaction are significantly improved. The system structure is compact, all adjustments are completed through software, hardware does not need to be changed, and the system is convenient to integrate into intelligent automobile central control, AR / VR headgear, interactive exhibition stands and other equipment, so that more rich and adaptive haptic experiences are provided for human-computer interaction. Under the premise of not changing the hardware structure, the system optimizes the phases and amplitudes of all channels of the ultrasonic array through the software algorithm of the upper computer, and cooperates with the hardware logic control of the FPGA controller to realize the dynamic control of the spatial resolution of the haptic focus and the generation of multiple foci.

The present invention relates to a lightweight method and system for detecting dangerous aerial drone movements. The method comprises: constructing a lightweight target detection network through a combination of depthwise separable convolution and dynamic resolution adjustment to locate drone targets in captured video frames; predicting and tracking the target's motion state through Kalman filtering in conjunction with drone target positioning; constructing a target memory library through a dynamic screening mechanism; and retrieving the most recent high-confidence feature from the target memory library to update the drone's target state when target association matching fails during tracking; and analyzing the drone's motion characteristics to detect dangerous drone movements. The present invention addresses the problems of high computing resource requirements, insufficient real-time performance, and poor generalization in the prior art.

This invention discloses a visual-tactile sensor based on high-density feature design and its image generation and tracking method. Addressing the problems of matching ambiguity and low tracking density caused by homogenized marker points in existing technologies, this invention first generates a globally unique feature pattern through an iterative algorithm with dual constraints in the spatial and color domains. Then, it utilizes adaptive field-of-view adjustment to compensate for feature distribution under different camera positions. This pattern is embedded in the sensor's tactile perception layer. During tracking, this invention employs a two-level cascaded matching logic, combined with reference frame modeling, dynamic resolution mechanisms, and mean normalization processing, to achieve displacement capture from pixel-level coarse matching to sub-pixel-level super-resolution. This invention significantly improves the sensor's displacement field tracking density, effectively eliminates feature confusion under large deformations, and enables high-precision perception at the full-field pixel level, providing a reliable high-density tactile feedback foundation for the refined operation of robots.

A video source device for wireless display sharing, comprising: an encoder designed to dynamically switch between encoding a full video frame into a first bitstream with a first resolution and a video subframe into a second bitstream with a second resolution, the second resolution being higher than the first resolution; a processing circuit arrangement designed to decide between encoding the full video frame and the video subframe based on an amount of available wireless transmission bandwidth, a number of pixels in a modified area of ​​the full video frame, spatial complexity of a modified area of ​​the full video frame, temporal complexity of a modified area of ​​the full video frame, and / or an area-change category of the full video frame; and a transmitter designed to wirelessly transmit the first bitstream and the second bitstream to a video sink device.

The invention discloses an image classification method and device based on visual large model Token parallelization, and relates to the technical field of image classification, and the method comprises the steps: inputting a to-be-processed image into a dynamic resolution predictor module to determine the optimal resolution of the to-be-processed image when a main node receives an image processing request, and carrying out the prediction of the optimal resolution of the to-be-processed image; adjusting the size of the to-be-processed image based on the optimal resolution to obtain a size-adjusted image; when the optimal resolution is greater than a preset resolution value, determining a dynamic planning result by using a dynamic planning algorithm according to the waiting queue and image processing requirements; packaging the first request in the waiting queue into a plurality of token packets; and according to a dynamic planning result, respectively sending all token packets to different edge devices for processing. According to the invention, the image token number optimization device and the reasoning service planning device are arranged, so that the response speed and stability of the edge device cluster during image classification can be improved.

The invention discloses a progressive visual content generation system based on thinking chain reasoning, which belongs to the technical field of visual content generation, and comprises a thinking chain reasoning module for constructing a semantic-driven logical reasoning framework and converting an abstract concept input by a user into an executable visual generation instruction; the progressive generation module executes a staged content generation strategy, adopts a'global-local-optimization 'three-order generation normal form, and improves the efficiency on the premise of ensuring the generation quality through a dynamic resolution switching and resource self-adaptive allocation mechanism; and a multi-modal data processing module. Through the synergistic effect of the semantic-driven thinking chain reasoning framework and the knowledge graph, it is ensured that generated content strictly follows physical rules and professional specifications, logic contradictions and common learning errors are effectively avoided, a global-to-local progressive generation strategy is adopted, a dynamic resource allocation mechanism is combined, and the generation efficiency is improved. And the generation efficiency is remarkably improved while the visual quality is ensured.

The application discloses a real-time rendering optimization method based on ray tracing. The application combines the optimization of spatial data structure, ray beam tracing technology, adaptive sampling, denoising technology, light map precalculation, shader optimization, ray-scene intersection acceleration, light source clipping, importance sampling, hardware acceleration, software-hardware cooperation, machine learning prediction, hybrid rendering technology, asynchronous calculation and dynamic resolution adjustment and other technologies; thereby improving the calculation efficiency of ray tracing and reducing the demand for hardware performance, so that the real-time or near real-time rendering speed is realized without sacrificing the image quality.

The embodiment of the invention relates to enhanced HDR image generation based on a gain map. A computing system for generating an enhanced high dynamic resolution (HDR) image based on a gain map receives an HDR image comprising a plurality of pixels, each pixel having one or more luminance values for each color component of a plurality of color components; applying at least a first tone mapping algorithm to each pixel in the HDR image; generating a gain map by using the HDR image and the SDR image; generating an enhanced HDR image based on the gain graph based on the SDR image and the gain graph; and generating an output based on the gain map-based enhanced HDR image.

The invention relates to a static graph mode graph generation system supporting dynamic resolution. The system comprises an original logic diagram construction component which is used for constructing an original logic diagram by tracking original calculation logic by executing a construction function of a diagram aiming at a first specific diagram generation task; the logic diagram optimization component is used for forming a first logic diagram according to the constructed original logic diagram; the node insertion component is used for inserting communication nodes and runtime control nodes required by distributed calculation according to the first logic diagram, and generating a first runtime logic diagram for multi-device collaboration; the operation plan generation component is used for generating a first runtime execution plan based on the data dependency relationship of the first runtime logic diagram; and the runtime initialization component is used for initializing the runtime instance of the first specific diagram according to the first runtime execution plan so as to complete dynamic allocation of computing resources and configuration of an execution environment.

The invention provides a dynamic resolution visual language model training method, a dynamic resolution visual language model detection method and dynamic resolution visual language model detection equipment. The dynamic resolution visual language model training method comprises the steps of obtaining a multi-modal document data set and a plurality of query text feature vectors; training a preset deep learning model to enable the deep learning model to detect each document image in the training process to generate a plurality of visual area maps with different resolutions corresponding to each document image, and extracting and fusing each visual area map to obtain an image feature vector, target loss is calculated based on the image feature vectors and the corresponding query text feature vectors to update model parameters of a deep learning model, and a trained deep learning model is obtained; and taking the deep learning model as a dynamic resolution visual language model for retrieving a document page corresponding to the query text feature vector from a plurality of document images. According to the method, the problem that the retrieval precision is limited due to the fact that complex layout and visual details in a document are ignored due to fixed-resolution input in a traditional method can be effectively solved.

The invention discloses a lightweight target detection method, and particularly relates to the technical field of target detection, and the method comprises the steps: firstly carrying out the preprocessing of an input image, and extracting low-calculation-amount and rich-semantic features through parallel convolution branch; then, an ICANeck lightweight trunk is adopted, and semantics are extracted in combination with divisible convolution and coordinate attention; constructing a strategy selection model, and switching between a fixed resolution-channel strategy and a dynamic resolution-channel strategy according to the target scale, the definition and the small target density to realize fine regulation and control of the resolution and the number of channels; the Stage 2-4 is subjected to abstraction deepening by using a progressive ICAblock; in the fusion stage, a proportional attention and channel alignment module is introduced to integrate scale features, and a lightweight detection head outputs a category and a frame; and sparse regularization and multi-task loss combined training is adopted, so that calculation and video memory overhead can be reduced, small target precision can be ensured, and edge real-time deployment can be adapted.

The application discloses a lightweight target detection method, and particularly relates to the technical field of target detection. Firstly, the input image is preprocessed, and low-computational and rich-semantic features are extracted through a parallel convolution branch. Then, an ICAneck lightweight backbone is adopted, and a separable convolution and a coordinate attention are combined to extract semantics. A strategy selection model is constructed, and according to the target scale, the definition and the small target density, switching is carried out between the fixed and dynamic resolution-channel strategies, so that the resolution and the number of channels are finely controlled. In stage 2-4, an incremental ICAblock is used for deepening abstraction. In the fusion stage, a proportional attention and a channel alignment module are introduced to integrate scale features, and a lightweight detection head is used to output the category and the frame. Through joint training of sparse regularization and multi-task loss, the calculation and the memory consumption are reduced, the accuracy of small targets is ensured, and the edge real-time deployment is adapted.

The invention discloses an image processing method and device, a storage medium and a processor. According to the scheme, an original image of the surrounding environment of a vehicle is collected, downsampling processing is carried out on the original image, and a preview is generated; segmenting the preview into a plurality of image areas, and calculating image entropies of all pixels in each image area; determining the region type of the corresponding region in the original image according to the image entropy of each image region; the region type comprises a plurality of salient regions of different grades, and the grades of the salient regions are in positive correlation with the image entropy; adjusting the resolution of a corresponding region in the original image according to the region type, and outputting the adjusted image as a target image; the resolution ratio of the corresponding region is positively correlated with the grade of the salient region corresponding to the corresponding region. Differentiation processing is carried out on the image through a dynamic resolution control mechanism, the computing power resource utilization rate of the intelligent driving system is improved, and the operation cost of the intelligent driving system is remarkably reduced.