Data backflow method, system, equipment and medium

Abstract

The invention provides a data backflow method, system and device and a medium, and belongs to the technical field of artificial intelligence, the method is applied to a core side management node, and network connections used for data backflow are pre-established between the core side management node and application nodes deployed in different geographic areas; the method comprises the following steps: sending artificial intelligence (AI) capability data to each application node, wherein the AI capability data is used for deploying AI capabilities at the application nodes; and sending a backflow task message to each application node, the backflow task message being used for indicating the application node to send a set of reasoning results obtained by processing image data by itself through the AI capability, so that the reasoning results of each application node are returned to the core side management node for storage.

Description

Technical Field

[0001] The present application relates to the field of artificial intelligence technology, and in particular to a data reflow method, system, device and medium. Background Art

[0002] With the rapid development of artificial intelligence (AI) technology and cloud computing technology, a two-level network architecture consisting of a core-side management node and different application nodes (which are usually deployed in different geographical areas) has been widely used.

[0003] However, since the inference results of each application node are generated, stored and used locally, that is, the inference results of each application node are circulated within the private domain of the relevant application node, the interoperability and application of data are greatly restricted, which makes it difficult to meet the data application needs of business scenarios such as AI capability optimization. Summary of the invention

[0004] In view of the above problems, embodiments of the present application provide a data reflow method, system, device and medium to overcome the above problems or at least partially solve the above problems.

[0005] In a first aspect of an embodiment of the present application, a data backflow method is provided, which is applied to a core-side management node, wherein a network connection for data backflow is pre-established between the core-side management node and application nodes deployed in different geographical areas; the method comprises: Sending artificial intelligence (AI) capability data to each of the application nodes, where the AI ​​capability data is used to deploy AI capabilities on the application nodes; A backflow task message is sent to each of the application nodes, where the backflow task message is used to instruct the application node to send a set of inference results obtained by processing the image data through the AI ​​capability, so that the inference results of each of the application nodes are backflowed to the core-side management node for storage.

[0006] As a possible implementation manner, sending a backflow task message to each of the application nodes includes: Sending a reflow task message carrying reflow strategy information to each of the application nodes, wherein the reflow strategy information is used to indicate the reasoning result of the required reflow; The reflux strategy information includes: at least one of a reflux ratio, a reflux upper limit, a geographical area, an AI capability interface, a reflux task start time, and a reflux task deadline.

[0007] As a possible implementation, the network connection is a CN2-BD network connection, and the set of inference results includes: the alarm data and / or non-alarm data obtained by the application node processing the image data collected in one or more scenarios through the AI ​​capability; The method further comprises: For the returned alarm data, use the alarm data to perform AI capability training for the one or more scenarios, and / or verify user feedback data related to false alarms; For the reflowed non-alarm data, the non-alarm data is used to perform AI capability training for the one or more scenarios, and / or perform analysis related to algorithm missed detections.

[0008] As a possible implementation, the method further includes: Setting access rights for the set of inference results stored in the core-side management node, wherein the access rights are used to deny data access or processing operations by unauthorized users; The set of inference results stored in the core-side management node records access to and operation logs of the data.

[0009] A second aspect of an embodiment of the present application provides a data backflow method, which is applied to an application node deployed in a single geographical area, wherein a network connection for data backflow is pre-established between the application node and a core-side management node; the method comprises: Receiving artificial intelligence (AI) capability data sent by the core-side management node, where the AI ​​capability data is used to deploy AI capabilities on the application node; Processing the multiple image data collected by the application node through the AI ​​capability to obtain a set of inference results; In response to receiving the backflow task message sent by the core-side management node, the set of inference results is sent to the core-side management node, so that the inference results of the application node are backflowed to the core-side management node for storage.

[0010] As a possible implementation manner, the reflow task message carries reflow strategy information, and the reflow strategy information includes: at least one of a reflow ratio, a reflow upper limit, a geographical area, an AI capability interface, a reflow task start time, and a reflow task deadline; The sending the set of inference results to the core-side management node includes: Filtering the inference results in the set of inference results according to the reflow strategy information to obtain a set of inference results for required reflow; The set of inference results required to be refluxed is sent to the core-side management node through the Kafka message middleware.

[0011] As a possible implementation manner, the network connection is a CN2-BD network connection; the processing of the multiple image data collected by the application node by the AI ​​capability to obtain a set of inference results includes: Processing the image data collected by the application node in one or more scenes by using the AI ​​capability, and obtaining the warning data and / or non-warning data associated with the one or more scenes as a set of the reasoning results; Among them, the alarm data in the set of reasoning results is used by the core side management node to perform AI capability training for the one or more scenarios, and / or to verify user feedback data related to false alarms; the non-alarm data in the set of reasoning results is used by the core side management node to perform AI capability training for the one or more scenarios, and / or to perform analysis related to algorithm missed detections.

[0012] According to a third aspect of an embodiment of the present application, a data backflow system is provided, the system comprising a core-side management node and application nodes deployed in different geographical areas, wherein a network connection for data backflow is pre-established between the core-side management node and each of the application nodes, wherein: The core-side management node is used to send artificial intelligence AI capability data to each of the application nodes, and the AI ​​capability data is used to deploy AI capabilities on the application nodes; Each of the application nodes is used to process the multiple image data collected by the application node through the AI ​​capability to obtain a set of inference results; The core-side management node is further used to send a backflow task message to each of the application nodes; Each of the application nodes is also used to send the set of reasoning results to the core side management node in response to receiving a backflow task message sent by the core side management node, so that the reasoning results of the application node flow back to the core side management node for storage.

[0013] According to a fourth aspect of an embodiment of the present application, an electronic device is provided, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the steps of the data reflow method described in the first aspect or the second aspect are implemented.

[0014] In a fifth aspect of an embodiment of the present application, a computer-readable storage medium is provided, on which a computer program / instruction is stored. When the computer program / instruction is executed by a processor, the steps of the data reflow method described in the first aspect or the second aspect are implemented.

[0015] The embodiments of the present application include the following advantages: a network connection for data backflow is pre-established between the core-side management node and the application node to ensure network intercommunication between the core-side management node and the application node, and a separate network connection is established for data backflow, which is also beneficial to improving the security of related transmission data during data backflow; on this basis, the core-side management node first sends AI capability data to the application node to implement the deployment of AI capability, and then sends a backflow task message to the application node to instruct the application node to send a set of inference results obtained by itself; through the aforementioned communication interaction process, the inference results of each application node will flow back to the core-side management node for unified storage, thereby breaking the technical barrier that the inference results of the application node cannot leave the private domain, realizing data intercommunication and full utilization of the inference results of each application node, thereby meeting the data application needs of business scenarios such as AI capability optimization. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings required for use in the description of the embodiments of the present application will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative labor.

[0017] Figure 1 is an implementation flow chart of a data reflow method in an embodiment of the present application; Figure 2 is a schematic diagram of an implementation process of a data reflow method in an embodiment of the present application; Figure 3 is a schematic diagram of a data backflow related component in an embodiment of the present application; Figure 4 is an implementation flow chart of another data reflow method in an embodiment of the present application; Figure 5 is a structural schematic diagram of a data backflow system in an embodiment of the present application; Figure 6 is a structural schematic diagram of a data backflow device in an embodiment of the present application; Figure 7 is a structural schematic diagram of another data backflow device in an embodiment of the present application; Figure 8 It is a schematic diagram of an electronic device in an embodiment of the present application. DETAILED DESCRIPTION

[0018] To facilitate understanding of the technical solutions provided by the present application, the main technical concepts involved in the embodiments of the present application are briefly described below.

[0019] Computer Vision (CV) algorithm: It is an important branch of artificial intelligence, focusing on enabling computers to understand and interpret visual information. Computer vision algorithms can simulate the human visual system and perform a variety of tasks, such as image recognition, object detection, scene reconstruction, image segmentation, image processing, etc.

[0020] Kafka message middleware: It is a high-throughput distributed publish-subscribe message system that can process large amounts of data and support offline data processing and real-time data processing.

[0021] CN2-BD: CN2 is a high-quality, high-performance bearer network that can provide users with faster, more stable and secure network connection services. It supports the integrated application of multiple services such as data, voice, and video, laying a solid foundation for the development of Next Generation Network (NGN) services. CN2-BD is a network subdivided by CN2 in the field of big data.

[0022] Data reflow: refers to the process of transferring the inference results obtained by the end-side algorithm from the cloud platform to a place closer to data generation or data application.

[0023] Under the two-level network architecture, in order to ensure the privacy and security of data (such as government data, AI scenario data, etc.), the inference results of each application node are generated, stored and used locally, that is, the inference results of each application node are circulated within the private domain of the relevant application node (such as the private network of the relevant business company), resulting in significant restrictions on the interoperability and application of data, and thus it is difficult to meet the data application needs of business scenarios such as AI capability optimization; for example, since the inference results of each application node cannot leave the private domain, it is difficult for users to obtain and use the inference results of each application node to optimize AI capabilities, and it is also difficult to conduct analysis related to algorithm omissions and false alarms based on the inference results of each application node.

[0024] In response to the problems existing in the above-mentioned related technologies, the embodiments of the present application propose a data backflow method, system, device and medium, by pre-establishing a network connection for data backflow between the core-side management node and the application node, and designing a communication interaction process related to data backflow, so that the reasoning results of each application node can be returned to the core-side management node for unified storage, thereby breaking the technical barrier that the reasoning results of the application nodes cannot leave the private domain, realizing data interoperability and full utilization of the reasoning results of each application node, thereby meeting the data application needs of business scenarios such as AI capability optimization.

[0025] In conjunction with the accompanying drawings, a data reflow method, system, device and medium provided by the embodiments of the present application are described in detail below through some embodiments and their application scenarios.

[0026] First, refer to Figure 1 As shown, it is an implementation flow chart of a data backflow method provided in an embodiment of the present application. The data backflow method is applied to a core-side management node. A network connection for data backflow is pre-established between the core-side management node and application nodes deployed in different geographical areas. The method comprises the following steps: Step S11: Send artificial intelligence AI capability data to each of the application nodes, where the AI ​​capability data is used to deploy AI capabilities on the application nodes.

[0027] Among them, the core-side management node and the application node constitute a two-level network architecture. The core-side management node is deployed with algorithm management services (such as algorithm warehouse management services), which can be used for training AI capabilities (i.e., CV algorithm capabilities) and deploying AI capabilities for application nodes; the application node (or it can be called a business-side inference node) is deployed with algorithm inference services, which can be used to perform inference tasks, that is, collect image data related to the inference task, and process the image data through AI capabilities to obtain inference results; the AI ​​capability may include CV algorithm capabilities suitable for one or more scenarios (such as CV large model algorithm, CV small model algorithm, etc.), which is used to process data collected by end-side devices (cameras, sensors, etc.) in different scenarios (such as image recognition, target detection and image classification, etc.) to obtain inference results.

[0028] In the specific implementation, based on the pre-established network connection for data backflow to achieve network connectivity between the core-side management node and the application node, the core-side management node can use data transmission protocols such as Hypertext Transfer Protocol (HTTP) to send AI capability data (such as copy data of the CV algorithm model, etc.) to the application node for AI capability deployment, thereby supporting the application node to call the corresponding AI capabilities to identify, parse and classify image data such as pictures and videos collected by the end-side device, and then obtain relevant reasoning results such as target detection results and image classification results.

[0029] Step S12: Send a backflow task message to each of the application nodes, where the backflow task message is used to instruct the application node to send a set of inference results obtained by processing image data through the AI ​​capability, so that the inference results of each of the application nodes flow back to the core-side management node for storage.

[0030] In specific implementation, the core-side management node creates a reflow task and can use data transmission protocols such as HTTP to send the reflow task message corresponding to the reflow task to each of the application nodes (or to some application nodes that need reflow data) to implement the reflow task distribution.

[0031] Each of the application nodes, upon receiving the backflow task message sent by the core-side management node, executes the backflow task, that is, sends a set of inference results obtained by the deployed AI capabilities to the core-side management node, thereby each of the application nodes can return the inference results determined by each node to the core-side management node for storage (for example, return to the distributed file system Ceph of the core-side management node for storage) by executing the backflow task. Optionally, the application node can use the kafka message middleware to send the set of inference results.

[0032] It is understandable that by implementing data reflow for the inference results of each of the application nodes, at least the following optimization effects can be obtained: 1. More comprehensive data access: Data reflow allows users to access more comprehensive data sets, that is, data from different application nodes that are widely distributed and related to different scenarios. This helps the algorithm obtain richer features and information during training, thereby improving its generalization ability and accuracy.

[0033] 2. Data privacy protection: For some application scenarios with high requirements for data privacy, by taking the relevant data from the private network (i.e., private domain) where the application node is located and returning it to the core management node for storage, use, and management through a pre-established network connection (i.e., a dedicated network channel for data return), the potential risk of data leakage in the public cloud can be avoided, ensuring the quality and security of data processed by the algorithm, thereby indirectly improving the reliability and accuracy of the algorithm. In this way, based on data return, this application can achieve better data access rights control for data with high privacy requirements (such as government data, etc.), thereby achieving data interoperability and full utilization on the premise of meeting its privacy requirements.

[0034] 3. Customized data processing: After the data flows back, users can process and optimize the data in more detail according to their own needs, such as data cleaning, feature engineering, etc. These customized processes can further improve the training effect of the algorithm model.

[0035] 4. Reduce costs: In the long run, returning the inference results of each application node to the core management node for storage will help reduce the data storage and cost of the application nodes, especially when the amount of data is extremely large. The cost reduction can allow users to invest more resources in the research and development and optimization of algorithms (i.e. AI capabilities), which is conducive to improving algorithm accuracy.

[0036] 5. Promote algorithm iterative optimization: Data reflux enables users to iterate and optimize algorithms more easily. That is, by continuously adjusting the algorithm using inference results based on real data, the performance and accuracy of the algorithm can be gradually improved, which helps to improve the recognition rate of the algorithm in different scenarios.

[0037] By adopting the technical solution of the embodiment of the present application, a network connection for data backflow is pre-established between the core-side management node and the application node to ensure network intercommunication between the core-side management node and the application node, and a separate network connection is established for data backflow, which is also beneficial to improve the security of related transmission data during data backflow; on this basis, the core-side management node first sends AI capability data to the application node to realize the deployment of AI capability, and then sends a backflow task message to the application node to instruct the application node to send a set of inference results obtained by itself; through the aforementioned communication interaction process, the inference results of each application node will be returned to the core-side management node for unified storage, thereby breaking the technical barrier that the inference results of the application node cannot leave the private domain, realizing data intercommunication and full utilization of the inference results of each application node, thereby meeting the data application needs of business scenarios such as AI capability optimization.

[0038] As a possible implementation manner, sending a backflow task message to each of the application nodes includes: Sending a reflow task message carrying reflow strategy information to each of the application nodes, wherein the reflow strategy information is used to indicate the reasoning result of the required reflow; The reflux strategy information includes: at least one of a reflux ratio, a reflux upper limit, a geographical area, an AI capability interface, a reflux task start time, and a reflux task deadline.

[0039] In the specific implementation, the core-side management node uses the reflow strategy information carried in the reflow task message to instruct the application node on the inference results required for reflow, thereby achieving accurate reflow of specific inference results (such as inference results related to a specific time, a specific application node, or a specific scenario), allowing users to flexibly obtain the required inference results according to actual needs.

[0040] For example, based on the indication of the reflow ratio or the reflow upper limit, the application node may select the inference results that meet the corresponding quantity or meet the corresponding ratio from the previously determined and temporarily stored inference results for reflow.

[0041] For another example, based on the geographical area (such as a province, city, etc.), the application node may determine whether it is necessary to return its own reasoning results to the core-side management node.

[0042] For another example, based on the AI ​​capability interface, the application node may know which AI capability interfaces correspond to the AI ​​capabilities that need to be used to obtain inference results that are fed back to the core-side management node.

[0043] For another example, based on the reflow task start time and the reflow task deadline, the application node may know when to start and when to stop sending the inference result to the core-side management node.

[0044] As a possible implementation, the network connection is a CN2-BD network connection. By using CN2-BD to establish a network connection (ie, a network channel) dedicated to data backflow, the transmission quality and security of data related to data backflow can be further improved.

[0045] Taking the establishment of network connection between the core-side management node and the provincial-side application node as an example, the public Internet Protocol (IP) address of the core-side management node is configured, that is, the public IP address is mapped to the fixed domain name on the provincial side through the dynamic Domain Name System (DNS) service; the provincial-side CN2-BD address is configured, that is, the specific port is mapped to a model service in the intranet; after the configuration is completed, multiple rounds of port connectivity verification are performed between the core-side management node and the provincial-side application node. If the ports are interoperable, it can be determined that the CN2-BD network connection is successfully established.

[0046] As a possible implementation, the set of inference results includes: warning data and / or non-warning data obtained by the application node through processing the image data collected in one or more scenarios by the AI ​​capability; The method further comprises: For the returned alarm data, use the alarm data to train the AI ​​capability for the one or more scenarios, and / or verify the user feedback data related to the false alarm, such as verifying the alarm data related to the false alarm reported by the user; For the reflowed non-alarm data, the non-alarm data is used to perform AI capability training for the one or more scenarios, and / or perform analysis related to algorithm missed detections.

[0047] In the specific implementation, the application node collects and stores the image data (such as pictures, videos) collected by the camera in one or more front-end scenarios (such as security, transportation, urban governance, etc.), and then uses the deployed AI capabilities to identify alarm data for the image data collected in different scenarios based on corresponding strategies (such as time strategy, scene strategy, and needle extraction frequency strategy, etc.), thereby classifying the image data collected in the one or more scenarios into alarm data and non-alarm data, that is, obtaining the inference data corresponding to each image data collected in the one or more scenarios.

[0048] Taking the AI ​​capability of face recognition as an example, the application node uses the AI ​​capability to compare the image data collected by the cameras in the park with the image data in the database of people in the park, and uses the image data of people who are not in the park as alarm data (in order to alarm the entry of people who are not in the park), and uses the image data of people who are in the park as non-alarm data.

[0049] Then, the application node can use the alarm data determined by itself, or the non-alarm data determined by itself, or the alarm data and non-alarm data determined by itself as a set of reasoning results according to the pre-negotiated agreement or the currently received reflow policy indication, and send the set of reasoning results to the core side management node to realize data reflow of the reasoning results.

[0050] After receiving a large amount of alarm data and / or non-alarm data from the return flow, the core management node can perform relevant analysis on missed detection or false alarms by the algorithm, or generate a data set file based on the return flow data to train the AI ​​capability (i.e., CV algorithm capability) to improve the recognition success rate of the algorithm in different scenarios. Among them, if the algorithm has a low recognition rate for data in a certain scenario, the return flow data corresponding to the scenario can be used to train the algorithm in a targeted manner to improve the algorithm's recognition accuracy for data in the scenario.

[0051] Optionally, data with correct inference results are filtered out from the reflowed alarm data and / or non-alarm data as a data set, and the algorithm is targeted for training and optimization using the data set, that is, by continuously performing multiple rounds of learning and algorithm improvement on the correct image data, the detection and classification accuracy of the CV algorithm and its applicability to different scenarios can be significantly improved.

[0052] In this embodiment, the present application is based on data backflow, which breaks the problem that the field data (i.e., the reasoning results) in the geographical area where the application node is located cannot leave the private domain, so that the alarm data (and / or non-alarm data) can flow back to the core AI capability production node (i.e., the core side management node), thereby providing real data sets of different scenarios for algorithm training, so that the core side management node can use the real data of different scenarios to analyze and optimize the training of AI capabilities, thereby improving the accuracy of the algorithm.

[0053] As a possible implementation, the method further includes: Setting access rights for the set of inference results stored in the core-side management node, wherein the access rights are used to deny data access or processing operations by unauthorized users; The set of inference results stored in the core management node records data access and operation logs for security auditing and tracing.

[0054] Reference Figure 2 The schematic diagram of the implementation process of the data backflow method shown in the figure takes the application nodes deployed in different geographical areas as provincial-side application nodes deployed in different provinces as an example to illustrate the implementation process of the data backflow method provided in the embodiment of the present application, and the implementation process includes: Step 1: The provincial application nodes deploy cameras to collect image data, and the core management nodes manage AI capability data.

[0055] In this step, refer to Figure 3 The schematic diagram of data backflow related components shown in the figure shows that a data collection component is deployed on the provincial application node. The data collection component uses data collection tools such as Flume to collect image data, and uses the programming framework of the distributed computing program (mapreduce) to clean the collected data, and then stores the relevant data in the Ceph of the provincial application node.

[0056] Step 2: The provincial-side application nodes and the core-side management nodes achieve network interconnection based on CN2-BD.

[0057] It should be noted that with the help of the CN2-BD network, the reliability and security of data during transmission can be ensured, and the CN2-BD network supports the core-side management node and the provincial-side application node to use HTTP, Hypertext Transfer Protocol Secure (HTTPS), File Transfer Protocol (FTP) and other data transmission protocols to transmit data backflow related signaling and data.

[0058] Step 3: The core-side management node sends AI capability data to the provincial-side application node.

[0059] Step 4: The provincial application node calls AI capabilities to process the collected image data and obtain alarm data.

[0060] Step 5: The core-side management node creates a backflow task and its corresponding backflow strategy, and sends the relevant information to the provincial-side application node via HTTP to instruct the provincial-side application node to collect and send alarm data.

[0061] In this step, the data collection component deployed in the provincial application node, after determining the province of data reflow and other reflow policy information, uses the Application Programming Interface (API) to call the log collector to collect the corresponding alarm data from the alarm data temporarily stored in the provincial application node.

[0062] The data transmission component deployed in the provincial application node sends the collected alarm data to the core management node through a message queue (such as Kafka message middleware). It can be understood that Kafka message middleware can realize asynchronous data transmission and load balancing, thereby improving the efficiency and stability of data transmission.

[0063] Step 6: The core-side management node applies the returned data to iterative optimization of AI capabilities to improve algorithm accuracy.

[0064] In the specific implementation, the data processing components deployed on the core-side management node use processing engines (such as Hadoop, Spark and other big data processing platforms) to clean the returned data (such as using data cleaning tools to identify and correct errors, duplicates, outliers, etc. in the data to improve data quality), convert (such as using data conversion tools to convert data from JSON format to XML format to meet the data interface requirements of different systems) and aggregate and other data processing to meet the subsequent application requirements of the returned data.

[0065] The data storage component deployed on the core-side management node then uses a relational database (such as MySQL), a non-relational database (such as MongoDB, Redis), or a data warehouse (such as Hive, Greenplum), etc. to store the processed data, and can configure data backup and recovery related functions during storage to prevent data loss or damage.

[0066] The data security component deployed on the core-side management node will then set strict access permissions for the stored data to ensure that only authorized users can access and process the data, and record data access and operation logs for security auditing and tracing.

[0067] Optionally, the core-side management node and the provincial-side application node may also deploy other components related to technical compatibility to ensure compatibility in terms of data format, interface protocol, etc.

[0068] In the above example, this application implements two-level data reflux in the cloud AI industry, which can meet the application requirements of inference results of application nodes in different scenarios (such as medical, social governance, transportation, security, etc.); thus, this application builds a complete set of data reflux implementation plans for cloud services and products, which can realize the "logical reflux" of local and municipal data in an intensive mode, thereby improving data governance efficiency and algorithm performance.

[0069] It should be noted that in addition to being used as a training data set to optimize algorithm models, the returned high-quality alarm data can also be used to enrich the data warehouse. For example, data reflow can be used to support the creation of a unified data management center, thereby providing basic data for the subsequent expansion of AI capabilities.

[0070] The following takes the provincial-side application node of Province A as an example to illustrate the implementation process of the data backflow method provided in the embodiment of the present application.

[0071] First, the CN2-BD network between the core-side management node and the application node on the provincial side of Province A is opened to achieve signaling intercommunication of the two-level architecture.

[0072] Then, the core-side management node sends the AI ​​capability data to the application node on the provincial side of Province A via HTTP.

[0073] After the application node on the provincial side of Province A deploys the inference service based on the AI ​​capability data, it collects the image data of the camera in the scenarios corresponding to different algorithm capabilities by executing the corresponding inference tasks.

[0074] The application node on the provincial side of Province A uses the local inference service to call AI capabilities to parse, identify and detect the collected on-site image data, and output alarm data (video or picture). The processed alarm data is stored locally.

[0075] The core-side management node sends a reflux task message carrying reflux policy information. The reflux policy information includes: specific province (such as Province A), AI capability interface, reflux upper limit, reflux ratio, reflux task start and end time, etc.

[0076] The application node on the provincial side of Province A then flows the alarm data stored in the application node on the provincial side of Province A back to the core-side management node through the Kafka message middleware. This enables the data of the application node on the provincial side of Province A to flow back to the nodes for AI capability production and training, enabling it to improve the recognition accuracy of AI capabilities through real customer site data, thus breaking the limitation of the privatized product solution that data does not support customer site data leaving the private domain.

[0077] Based on the above example, this application proposes a data backflow method for a cloud-based two-level network architecture. The provincial-side application nodes and the core-side management nodes achieve network interconnection based on the pre-established CN2-BD network connection, thereby providing data security with the help of a network communication environment dedicated to data backflow.

[0078] At the same time, when the network is connected, the ability to flow back the inference results from the provincial application nodes to the core management nodes (i.e., the data backflow capability under the two-level network architecture) is realized, so that the algorithm can access a more comprehensive data set, which makes it easier for users to iterate and optimize the algorithm, and improve the recognition rate of the algorithm in different scenarios, thus providing an effective optimization method for CV algorithms in the AI ​​industry.

[0079] In addition, by returning data to the core-side management node, a large amount of high-quality forward AI algorithm data can be provided to the data warehouse of the core-side management node, while also providing the core-side management node with the ability to develop more business scenarios.

[0080] Second, refer to Figure 4 As shown, it is an implementation flow chart of another data backflow method provided in an embodiment of the present application, the data backflow method is applied to application nodes deployed in a single geographical area, and a network connection for data backflow is pre-established between the application node and the core-side management node; the method comprises the following steps: Step S21: receiving artificial intelligence (AI) capability data sent by the core-side management node, where the AI ​​capability data is used to deploy AI capabilities on the application node.

[0081] Step S22: Process the multiple image data collected by the application node through the AI ​​capability to obtain a set of inference results.

[0082] Step S23: In response to receiving the backflow task message sent by the core-side management node, sending the set of inference results to the core-side management node, so that the inference results of the application node flow back to the core-side management node for storage.

[0083] By adopting the technical solution of the embodiment of the present application, a network connection for data backflow is pre-established between the core-side management node and the application node to ensure network intercommunication between the core-side management node and the application node, and a separate network connection is established for data backflow, which is also beneficial to improve the security of related transmission data during data backflow; on this basis, the core-side management node first sends AI capability data to the application node to realize the deployment of AI capability, and then sends a backflow task message to the application node to instruct the application node to send a set of inference results obtained by itself; through the aforementioned communication interaction process, the inference results of each application node will be returned to the core-side management node for unified storage, thereby breaking the technical barrier that the inference results of the application node cannot leave the private domain, realizing data intercommunication and full utilization of the inference results of each application node, thereby meeting the data application needs of business scenarios such as AI capability optimization.

[0084] As a possible implementation manner, the reflow task message carries reflow strategy information, and the reflow strategy information includes: at least one of a reflow ratio, a reflow upper limit, a geographical area, an AI capability interface, a reflow task start time, and a reflow task deadline; The sending the set of inference results to the core-side management node includes: Filtering the inference results in the set of inference results according to the reflow strategy information to obtain a set of inference results for required reflow; The set of inference results required to be refluxed is sent to the core-side management node through the Kafka message middleware.

[0085] As a possible implementation manner, the network connection is a CN2-BD network connection; the processing of the multiple image data collected by the application node by the AI ​​capability to obtain a set of inference results includes: Processing the image data collected by the application node in one or more scenes by using the AI ​​capability, and obtaining the warning data and / or non-warning data associated with the one or more scenes as a set of the reasoning results; Among them, the alarm data in the set of reasoning results is used by the core side management node to perform AI capability training for the one or more scenarios, and / or to verify user feedback data related to false alarms; the non-alarm data in the set of reasoning results is used by the core side management node to perform AI capability training for the one or more scenarios, and / or to perform analysis related to algorithm missed detections.

[0086] For the method embodiments, for the sake of simplicity, they are all described as a series of action combinations, but those skilled in the art should be aware that the embodiments of the present application are not limited by the order of the actions described, because according to the embodiments of the present application, some steps can be performed in other orders or simultaneously. Secondly, those skilled in the art should also be aware that the embodiments described in the specification are all preferred embodiments, and the actions involved are not necessarily required by the embodiments of the present application.

[0087] Thirdly, Figure 5 : is a structural diagram of a data backflow system according to an embodiment of the present application, wherein the system includes a core-side management node and application nodes deployed in different geographical areas, and a network connection for data backflow is pre-established between the core-side management node and each of the application nodes, wherein: The core-side management node is used to send artificial intelligence AI capability data to each of the application nodes, and the AI ​​capability data is used to deploy AI capabilities on the application nodes; Each of the application nodes is used to process the multiple image data collected by the application node through the AI ​​capability to obtain a set of inference results; The core-side management node is further used to send a backflow task message to each of the application nodes; Each of the application nodes is also used to send the set of reasoning results to the core side management node in response to receiving a backflow task message sent by the core side management node, so that the reasoning results of the application node flow back to the core side management node for storage.

[0088] By adopting the technical solution of the embodiment of the present application, a network connection for data backflow is pre-established between the core-side management node and the application node to ensure network intercommunication between the core-side management node and the application node, and a separate network connection is established for data backflow, which is also beneficial to improve the security of related transmission data during data backflow; on this basis, the core-side management node first sends AI capability data to the application node to realize the deployment of AI capability, and then sends a backflow task message to the application node to instruct the application node to send a set of inference results obtained by itself; through the aforementioned communication interaction process, the inference results of each application node will be returned to the core-side management node for unified storage, thereby breaking the technical barrier that the inference results of the application node cannot leave the private domain, realizing data intercommunication and full utilization of the inference results of each application node, thereby meeting the data application needs of business scenarios such as AI capability optimization.

[0089] Optionally, the core-side management node is further used to send a reflow task message carrying reflow policy information to each of the application nodes, where the reflow policy information is used to indicate the reasoning result of the required reflow; Each of the application nodes is further used to filter the inference results in the set of inference results according to the reflow strategy information to obtain a set of inference results required for reflow; Each of the application nodes is further used to send the set of inference results required to be refluxed to the core-side management node through the Kafka message middleware; The reflux strategy information includes: at least one of a reflux ratio, a reflux upper limit, a geographical area, an AI capability interface, a reflux task start time, and a reflux task deadline.

[0090] Optionally, the network connection is a CN2-BD network connection; Each of the application nodes is further configured to process the image data collected by the application node in one or more scenarios through the AI ​​capability, and send the obtained alarm data and / or non-alarm data associated with the one or more scenarios as a collection of the inference results to the core-side management node; The core-side management node is also used to use the reflowed alarm data to perform AI capability training for the one or more scenarios, and / or to verify user feedback data related to false alarms; and to use the reflowed non-alarm data to perform AI capability training for the one or more scenarios, and / or to perform analysis related to algorithm missed detections.

[0091] Optionally, the core side management node is also used to set access rights for the set of inference results stored by the core side management node, and the access rights are used to deny data access or processing operations by unauthorized users; and to record data access and operation logs for the set of inference results stored by the core side management node.

[0092] Fourthly, Figure 6 : is a structural schematic diagram of a data backflow device according to an embodiment of the present application, the device is applied to a core-side management node, and a network connection for data backflow is pre-established between the core-side management node and application nodes deployed in different geographical areas; the device includes: A capability deployment module 610 is used to send artificial intelligence AI capability data to each of the application nodes, where the AI ​​capability data is used to deploy AI capabilities on the application nodes; The message sending module 620 is used to send a backflow task message to each of the application nodes, and the backflow task message is used to instruct the application node to send a set of inference results obtained by processing the image data through the AI ​​capability, so that the inference results of each of the application nodes are backflowed to the core-side management node for storage.

[0093] By adopting the technical solution of the embodiment of the present application, a network connection for data backflow is pre-established between the core-side management node and the application node to ensure network intercommunication between the core-side management node and the application node, and a separate network connection is established for data backflow, which is also beneficial to improve the security of related transmission data during data backflow; on this basis, the core-side management node first sends AI capability data to the application node to realize the deployment of AI capability, and then sends a backflow task message to the application node to instruct the application node to send a set of inference results obtained by itself; through the aforementioned communication interaction process, the inference results of each application node will be returned to the core-side management node for unified storage, thereby breaking the technical barrier that the inference results of the application node cannot leave the private domain, realizing data intercommunication and full utilization of the inference results of each application node, thereby meeting the data application needs of business scenarios such as AI capability optimization.

[0094] Optionally, the message sending module 620 is further used to send a reflow task message carrying reflow strategy information to each of the application nodes, where the reflow strategy information is used to indicate the reasoning result of the required reflow; The reflux strategy information includes: at least one of a reflux ratio, a reflux upper limit, a geographical area, an AI capability interface, a reflux task start time, and a reflux task deadline.

[0095] Optionally, the network connection is a CN2-BD network connection, and the set of inference results includes: warning data and / or non-warning data obtained by the application node processing image data collected in one or more scenarios through the AI ​​capability; The device also includes: A first processing module is used to use the reflowed alarm data to perform AI capability training for the one or more scenarios and / or verify user feedback data related to false alarms; The second processing module is used to use the non-alarm data for the reflowed non-alarm data to perform AI capability training for the one or more scenarios, and / or perform analysis related to algorithm missed detection.

[0096] Optionally, the device further comprises: A permission setting module, used to set access rights for the set of inference results stored in the core-side management node, wherein the access rights are used to deny data access or processing operations by unauthorized users; A logging module is used to record access and operation logs of the set of inference results stored in the core-side management node.

[0097] Fifthly, Figure 7: is a structural schematic diagram of another data backflow device according to an embodiment of the present application, the device is applied to an application node deployed in a single geographical area, and a network connection for data backflow is pre-established between the application node and the core-side management node; the device includes: A data receiving module 710 is used to receive artificial intelligence AI capability data sent by the core-side management node, where the AI ​​capability data is used to deploy AI capabilities on the application node; A data reasoning module 720, configured to process the plurality of image data collected by the application node through the AI ​​capability to obtain a set of reasoning results; The data sending module 730 is used to send the set of inference results to the core side management node in response to receiving the backflow task message sent by the core side management node, so that the inference results of the application node flow back to the core side management node for storage.

[0098] By adopting the technical solution of the embodiment of the present application, a network connection for data backflow is pre-established between the core-side management node and the application node to ensure network intercommunication between the core-side management node and the application node, and a separate network connection is established for data backflow, which is also beneficial to improve the security of related transmission data during data backflow; on this basis, the core-side management node first sends AI capability data to the application node to realize the deployment of AI capability, and then sends a backflow task message to the application node to instruct the application node to send a set of inference results obtained by itself; through the aforementioned communication interaction process, the inference results of each application node will be returned to the core-side management node for unified storage, thereby breaking the technical barrier that the inference results of the application node cannot leave the private domain, realizing data intercommunication and full utilization of the inference results of each application node, thereby meeting the data application needs of business scenarios such as AI capability optimization.

[0099] Optionally, the reflow task message carries reflow strategy information, and the reflow strategy information includes: at least one of a reflow ratio, a reflow upper limit, a geographical area, an AI capability interface, a reflow task start time, and a reflow task deadline; The data sending module 730 is further configured to perform the following steps: Filtering the inference results in the set of inference results according to the reflow strategy information to obtain a set of inference results for required reflow; The set of inference results required to be refluxed is sent to the core-side management node through the Kafka message middleware.

[0100] Optionally, the network connection is a CN2-BD network connection; The data reasoning module 720 is further configured to process the image data collected by the application node in one or more scenarios through the AI ​​capability, and obtain the warning data and / or non-warning data associated with the one or more scenarios as a set of the reasoning results; Among them, the alarm data in the set of reasoning results is used by the core side management node to perform AI capability training for the one or more scenarios, and / or to verify user feedback data related to false alarms; the non-alarm data in the set of reasoning results is used by the core side management node to perform AI capability training for the one or more scenarios, and / or to perform analysis related to algorithm missed detections.

[0101] It should be noted that the device embodiment is similar to the method embodiment, so the description is relatively simple, and the relevant parts can be referred to the method embodiment.

[0102] The present application also provides an electronic device, referring to Figure 8 , Figure 8 Schematic diagram of an electronic device proposed in an embodiment of the present application. Figure 8 As shown, the electronic device 100 includes: a memory 110 and a processor 120. The memory 110 and the processor 120 are connected via a bus communication. A computer program is stored in the memory 110. The computer program can be run on the processor 120 to implement the steps in the data reflow method disclosed in the embodiment of the present application.

[0103] The embodiment of the present application also provides a computer-readable storage medium on which a computer program / instruction is stored. When the computer program / instruction is executed by a processor, the data reflow method disclosed in the embodiment of the present application is implemented.

[0104] The embodiment of the present application also provides a computer program product, including a computer program / instruction, which, when executed by a processor, implements the data reflow method disclosed in the embodiment of the present application.

[0105] The various embodiments in this specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the various embodiments can be referenced to each other.

[0106] Those skilled in the art will appreciate that the embodiments of the present application can be provided as methods, devices or computer program products. Therefore, the present application can adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment in combination with software and hardware. Moreover, the present application can adopt the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0107] The embodiments of the present application are described with reference to the flowcharts and / or block diagrams of the methods, systems, devices, storage media, and program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing terminal device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing terminal device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0108] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing terminal device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce a manufactured product including an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0109] These computer program instructions can also be loaded onto a computer or other programmable data processing terminal device so that a series of operating steps are executed on the computer or other programmable terminal device to produce a computer-implemented process, thereby providing instructions for executing on the computer or other programmable terminal device to implement the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0110] Although the preferred embodiments of the present application have been described, those skilled in the art may make additional changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the embodiments of the present application.

[0111] Finally, it should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or terminal device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or terminal device. In the absence of further restrictions, the elements defined by the sentence "including one..." do not exclude the existence of other identical elements in the process, method, article or terminal device including the elements.

[0112] The above is a detailed introduction to a data reflow method, system, device and medium provided by the present application. Specific examples are used in this article to illustrate the principles and implementation methods of the present application. The description of the above embodiments is only used to help understand the method of the present application and its core idea. At the same time, for a person skilled in the art, according to the idea of ​​the present application, there will be changes in the specific implementation method and application scope. In summary, the content of this specification should not be understood as a limitation on the present application.

Claims

1. A data reflow method, characterized in that: Applied to a core-side management node, a network connection for data backflow is pre-established between the core-side management node and application nodes deployed in different geographical areas; the method comprises: Sending artificial intelligence (AI) capability data to each of the application nodes, where the AI ​​capability data is used to deploy AI capabilities on the application nodes; A backflow task message is sent to each of the application nodes, where the backflow task message is used to instruct the application node to send a set of inference results obtained by processing the image data through the AI ​​capability, so that the inference results of each of the application nodes are backflowed to the core-side management node for storage.

2. The method according to claim 1, characterized in that Sending a backflow task message to each of the application nodes includes: Sending a reflow task message carrying reflow strategy information to each of the application nodes, wherein the reflow strategy information is used to indicate the reasoning result of the required reflow; The reflux strategy information includes: at least one of a reflux ratio, a reflux upper limit, a geographical area, an AI capability interface, a reflux task start time, and a reflux task deadline.

3. The method according to claim 1, characterized in that The network connection is a CN2-BD network connection, and the set of inference results includes: the alarm data and / or non-alarm data obtained by the application node through processing the image data collected in one or more scenarios through the AI ​​capability; The method further comprises: For the returned alarm data, use the alarm data to perform AI capability training for the one or more scenarios, and / or verify user feedback data related to false alarms; For the reflowed non-alarm data, the non-alarm data is used to perform AI capability training for the one or more scenarios, and / or perform analysis related to algorithm missed detections.

4. The method according to any one of claims 1 to 3, characterized in that: The method further comprises: Setting access rights for the set of inference results stored in the core-side management node, wherein the access rights are used to deny data access or processing operations by unauthorized users; The set of inference results stored in the core-side management node records access to and operation logs of the data.

5. A data reflow method, characterized in that: Applied to application nodes deployed in a single geographical area, a network connection for data backflow is pre-established between the application node and the core-side management node; the method comprises: Receiving artificial intelligence (AI) capability data sent by the core-side management node, where the AI ​​capability data is used to deploy AI capabilities on the application node; Processing the multiple image data collected by the application node through the AI ​​capability to obtain a set of inference results; In response to receiving the backflow task message sent by the core-side management node, the set of inference results is sent to the core-side management node, so that the inference results of the application node are backflowed to the core-side management node for storage.

6. The method according to claim 5, characterized in that The reflow task message carries reflow strategy information, and the reflow strategy information includes: at least one of a reflow ratio, a reflow upper limit, a geographical area, an AI capability interface, a reflow task start time, and a reflow task deadline; The sending the set of inference results to the core-side management node includes: Filtering the inference results in the set of inference results according to the reflow strategy information to obtain a set of inference results for required reflow; The set of inference results required to be refluxed is sent to the core-side management node through the Kafka message middleware.

7. The method according to claim 5 or 6, characterized in that: The network connection is a CN2-BD network connection; the multiple image data collected by the application node are processed by the AI ​​capability to obtain a set of inference results, including: Processing the image data collected by the application node in one or more scenes by using the AI ​​capability, and obtaining the warning data and / or non-warning data associated with the one or more scenes as a set of the reasoning results; Among them, the alarm data in the set of reasoning results is used by the core side management node to perform AI capability training for the one or more scenarios, and / or to verify user feedback data related to false alarms; the non-alarm data in the set of reasoning results is used by the core side management node to perform AI capability training for the one or more scenarios, and / or to perform analysis related to algorithm missed detections.

8. A data backflow system, characterized in that: The system includes a core-side management node and application nodes deployed in different geographical areas, and a network connection for data backflow is pre-established between the core-side management node and each of the application nodes, wherein: The core-side management node is used to send artificial intelligence AI capability data to each of the application nodes, and the AI ​​capability data is used to deploy AI capabilities on the application nodes; Each of the application nodes is used to process the multiple image data collected by the application node through the AI ​​capability to obtain a set of inference results; The core-side management node is further used to send a backflow task message to each of the application nodes; Each of the application nodes is also used to send the set of reasoning results to the core side management node in response to receiving a backflow task message sent by the core side management node, so that the reasoning results of the application node flow back to the core side management node for storage.

9. An electronic device comprising a memory, a processor and a computer program stored in the memory, characterized in that: The processor executes the computer program to implement the data reflow method according to any one of claims 1 to 4, or to implement the data reflow method according to any one of claims 5 to 7.

10. A computer-readable storage medium having a computer program / instruction stored thereon, characterized in that: When the computer program / instruction is executed by a processor, the data reflow method according to any one of claims 1 to 4 is implemented, or the data reflow method according to any one of claims 5 to 7 is implemented.

Images