WO2023065746A1 - 算法应用元生成方法、装置、电子设备、计算机程序产品及计算机可读存储介质 - Google Patents
算法应用元生成方法、装置、电子设备、计算机程序产品及计算机可读存储介质 Download PDFInfo
- Publication number
- WO2023065746A1 WO2023065746A1 PCT/CN2022/107174 CN2022107174W WO2023065746A1 WO 2023065746 A1 WO2023065746 A1 WO 2023065746A1 CN 2022107174 W CN2022107174 W CN 2022107174W WO 2023065746 A1 WO2023065746 A1 WO 2023065746A1
- Authority
- WO
- WIPO (PCT)
- Prior art keywords
- template
- target
- algorithm application
- configuration
- function
- Prior art date
Links
- 238000004422 calculation algorithm Methods 0.000 title claims abstract description 328
- 238000000034 method Methods 0.000 title claims abstract description 106
- 238000004590 computer program Methods 0.000 title claims abstract description 15
- 238000012545 processing Methods 0.000 claims abstract description 84
- 230000006870 function Effects 0.000 claims description 166
- 230000008569 process Effects 0.000 claims description 56
- 238000001514 detection method Methods 0.000 claims description 47
- 238000000605 extraction Methods 0.000 claims description 28
- 230000015654 memory Effects 0.000 claims description 12
- 230000004044 response Effects 0.000 claims 1
- 238000005516 engineering process Methods 0.000 description 21
- 238000011161 development Methods 0.000 description 15
- 238000010586 diagram Methods 0.000 description 9
- 238000013473 artificial intelligence Methods 0.000 description 8
- 238000003062 neural network model Methods 0.000 description 7
- 206010039203 Road traffic accident Diseases 0.000 description 5
- 238000004891 communication Methods 0.000 description 5
- 238000013527 convolutional neural network Methods 0.000 description 3
- 238000013515 script Methods 0.000 description 3
- 238000012549 training Methods 0.000 description 3
- 238000013461 design Methods 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 230000006872 improvement Effects 0.000 description 2
- 230000003993 interaction Effects 0.000 description 2
- 238000012706 support-vector machine Methods 0.000 description 2
- 230000003190 augmentative effect Effects 0.000 description 1
- 238000013475 authorization Methods 0.000 description 1
- 230000006399 behavior Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000013135 deep learning Methods 0.000 description 1
- 238000013136 deep learning model Methods 0.000 description 1
- 238000012217 deletion Methods 0.000 description 1
- 230000037430 deletion Effects 0.000 description 1
- 230000000977 initiatory effect Effects 0.000 description 1
- 238000012804 iterative process Methods 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000003058 natural language processing Methods 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 230000008520 organization Effects 0.000 description 1
- 230000008447 perception Effects 0.000 description 1
- 238000009877 rendering Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/217—Validation; Performance evaluation; Active pattern learning techniques
Definitions
- the present disclosure relates to computer vision technology, and in particular to a method, device, electronic equipment, computer program product, and computer-readable storage medium for generating an algorithm application element.
- Algorithm application unit is a software unit for intelligent analysis of images or videos, which can be applied to various business scenarios, such as traffic flow early warning scenarios, traffic accident road section early warning scenarios, etc.
- the generation process of the algorithm application element includes creating the algorithm application element and debugging the algorithm application element.
- Embodiments of the present disclosure at least provide a method, device, electronic device, computer program product, and computer-readable storage medium for generating algorithm application elements, which can improve the efficiency of generating and updating algorithm application elements.
- An embodiment of the present disclosure provides a method for generating an algorithm application element, including:
- the template configuration item and the target initial template are obtained;
- the target initial template is a modularized general processing flow template that realizes the target image processing function in a business scenario;
- the template configuration item is used according to Presetting target requirements, defining at least one functional module to be realized in the general processing flow template;
- the realization configuration information in the generation instruction obtain at least one preset target function realization module;
- the realization configuration information is the at least one Implementation information corresponding to the functional module;
- a target algorithm application element is generated based on the at least one target function realization module and the target algorithm application element template.
- An embodiment of the present disclosure provides an algorithm application element generation device, including:
- the obtaining part is configured to obtain template configuration items and target initial templates by receiving application meta-template creation instructions;
- the target initial templates are modularized general processing flow templates that implement target image processing functions in business scenarios;
- the The template configuration item is used to define at least one functional module to be implemented in the general processing flow template according to preset target requirements;
- the template creation part is configured to generate a target algorithm application meta-template based on the template configuration item and the target initial template;
- the generating part is configured to obtain at least one preset target function realization module according to the realization configuration information in the generation instruction when receiving the generation instruction for the target algorithm application meta-template; the realization configuration The information is the implementation information corresponding to the at least one function module; and the target algorithm application element is generated based on the at least one target function realization module and the target algorithm application element template.
- An embodiment of the present disclosure provides an electronic device, including:
- a memory configured to store executable algorithm application element generation instructions
- the processor When the processor is configured to execute the executable algorithm application element generation instruction stored in the memory, implement the algorithm application element generation method provided by the embodiment of the present disclosure.
- An embodiment of the present disclosure provides a computer-readable storage medium, storing executable algorithm application element generation instructions configured to cause a processor to implement the algorithm application element generation method provided by the embodiment of the present disclosure.
- An embodiment of the present disclosure provides a computer program product, including a computer program or an instruction.
- the computer program or instruction is executed by a processor, the above method for generating an algorithm application element is realized.
- the embodiments of the present disclosure have the following technical effects: the general processing flow for intelligent processing of images or videos in actual business scenarios is abstracted into the target initial template, and according to the preset target requirements of the actual business, the template configuration items are used in the target initial template Custom configuration of the modularized general processing flow to obtain the target algorithm application meta-template corresponding to the actual business requirements; apply the meta-template for the target algorithm, and further configure the configuration information to generate the target algorithm application for realizing the actual business needs Yuan.
- the redundant and repeated work of independently developing algorithm application elements each time according to different business requirements is reduced, thereby reducing the generation time of algorithm application elements and improving the generation efficiency of algorithm application elements.
- the algorithm application element template or the implementation configuration information can be modified according to the field data, so as to realize the rapid update and iteration of the algorithm application element, thereby improving the update efficiency of the algorithm application element.
- FIG. 1 is a schematic diagram of an optional architecture of an algorithm application unit generating system 100 provided by an embodiment of the present disclosure
- FIG. 2 is a first schematic flow diagram of a method for generating an algorithm application element provided by an embodiment of the present disclosure
- Fig. 3 is an optional schematic flowchart II of the method for generating an algorithm application element provided by an embodiment of the present disclosure
- Fig. 4 is an optional schematic flow diagram III of the algorithm application element generation method provided by the embodiment of the present disclosure.
- Fig. 5 is an optional schematic flowchart 4 of the algorithm application element generation method provided by the embodiment of the present disclosure.
- FIG. 6 is an optional schematic flowchart five of the method for generating an algorithm application element provided by an embodiment of the present disclosure
- FIG. 7 is a schematic flow diagram of generating algorithm application elements in an actual application scenario provided by an embodiment of the present disclosure.
- FIG. 8 is a schematic structural diagram of an algorithm application element generation device provided by an embodiment of the present disclosure.
- FIG. 9 is a schematic structural diagram of an electronic device provided by an embodiment of the present disclosure.
- AI Artificial Intelligence
- a comprehensive technique of computer science that attempts to understand the essence of intelligence and generate a new kind of intelligent machine that can respond in a similar way to human intelligence.
- Artificial intelligence is to study the design principles and implementation methods of various intelligent machines, so that the machines have the functions of perception, reasoning and decision-making.
- Artificial intelligence technology is a comprehensive subject with a wide range of design fields, including both hardware-level technology and software-level technology.
- Artificial intelligence basic technologies generally include technologies such as sensors, dedicated artificial intelligence chips, cloud computing, distributed storage, big data processing technology, operation/interaction systems, and mechatronics.
- Artificial intelligence software technology includes several major directions such as computer vision technology, speech processing technology, natural language processing technology, and machine learning/deep learning.
- Computer Vision Technology Computer is actually a science that studies how to make machines "see”. Further, it refers to machines that use cameras and computers instead of human eyes to identify, track and measure targets. Vision, and further image processing, so that the computer processing becomes an image that is more suitable for human observation or sent to the instrument for detection.
- Computer vision technology usually includes image processing, image recognition, image semantic understanding, image retrieval, OCR, video analysis, video semantic understanding, video content/behavior recognition, 3D object reconstruction, 3D technology, virtual reality, augmented reality, simultaneous positioning and maps It also includes common face recognition, fingerprint recognition and other biometric recognition technologies.
- Image/video analysis refers to the analysis and processing of images collected by image acquisition equipment, so as to determine whether a specific event has occurred in the image or video. For example, by identifying and processing the video, it is analyzed whether the vehicle in the video has a traffic accident, or whether the person in the video has fallen, etc.
- the algorithm application element is a software module for analyzing and processing images or videos, which can integrate algorithm models required for image or video analysis, as well as scheduling service components, message service components, and storage service components. Input the image or video in the specified scene into the algorithm application element, and the analysis result corresponding to the scene can be obtained. In some embodiments, it is also possible to continue to develop based on the algorithm application element, for example, continue to develop interface interaction functions, communication functions, etc. on the basis of the algorithm application element, so as to obtain a complete application software based on the algorithm application element.
- Algorithm application unit is a software unit for intelligent analysis of images or videos, which can be applied to various business scenarios, such as traffic flow early warning scenarios, traffic accident road section early warning scenarios, etc.
- the generation process of the algorithm application element includes creating the algorithm application element and debugging the algorithm application element.
- the development of the algorithm application element in the related technology is to redevelop the algorithm application element after each algorithm model is released.
- it is necessary to manually write each file of the algorithm application element such as the directory structure, code files, configuration files, document files, and script files (Makefile, etc.) of the algorithm application element. content, as well as manually writing codes in code files, etc., so that the development time and labor costs of generating algorithm application elements increase linearly, reducing the generation efficiency of algorithm application elements.
- Embodiments of the present disclosure provide a method, device, electronic device, computer program product, and computer-readable storage medium for generating algorithm application elements, which can improve the efficiency of generating and updating algorithm application elements.
- the exemplary application of the electronic equipment provided by the embodiments of the present disclosure is described below.
- the electronic equipment provided by the embodiments of the present disclosure can be implemented as a notebook computer, a tablet computer, a desktop computer, a set-top box, a mobile device (for example, a mobile phone, a portable music player, Various types of user terminals such as personal digital assistants, dedicated messaging devices, and portable game devices) can also be implemented as servers.
- an exemplary application in which an electronic device is implemented as a server will be explained.
- FIG. 1 is a schematic diagram of an optional architecture of an algorithm application element generating system 100 provided by an embodiment of the present disclosure.
- the terminal 400 is connected to the server 200 through the network 300.
- the network 300 may be a wide area network or a local area network, or a combination of the two.
- the terminal 400 belongs to a business scenario, such as an engineering developer at a project debugging site, the developer can use the terminal 400 to send an application meta template creation instruction to the server 200 according to actual business requirements;
- the server 200 may be a background server, which is used to obtain template configuration items and target initial templates by receiving application meta-template creation instructions; the target initial templates are modularized general processing flow templates that realize target image processing functions in business scenarios; templates The configuration item is used to define at least one functional module to be implemented in the general processing flow template according to the preset target requirements; based on the template configuration item and the target initial template, generate the target algorithm application meta-template; after receiving the target algorithm application meta-template
- at least one preset target function realization module is obtained according to the realization configuration information in the generation instruction; the realization configuration information is the realization information corresponding to at least one function module; based on at least one target function realization module and the target algorithm Application element template to generate target algorithm application elements.
- the server 200 is further configured to send the generated target algorithm application element to the terminal 400, so that the terminal 400 analyzes and processes the video or image in the business scene by executing the target algorithm application element, and obtains a processing result; and, according to the processing result Perform further operations such as configuration update on the target algorithm application element.
- the server 200 can be an independent physical server, or a server cluster or a distributed system composed of multiple physical servers, and can also provide cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network Cloud servers for basic cloud computing services such as cloud services, cloud communications, middleware services, domain name services, security services, CDN, and big data and artificial intelligence platforms.
- the terminal 400 may be a smart phone, a tablet computer, a notebook computer, a desktop computer, a smart speaker, a smart watch, etc., but is not limited thereto.
- the terminal and the server may be connected directly or indirectly through wired or wireless communication, which is not limited in this embodiment of the present disclosure.
- FIG. 2 is an optional schematic flowchart 1 of a method for generating an algorithm application element provided by an embodiment of the present disclosure, which will be described in conjunction with the steps shown in FIG. 2 .
- S101 Obtain template configuration items and target initial templates by receiving an application meta-template creation instruction;
- the target initial template is a modularized general processing flow template that realizes the target image processing function in a business scenario;
- the template configuration items are used according to preset
- the target requirement defines at least one functional module to be realized in the general processing flow template.
- the embodiments of the present disclosure are applicable to the generation and debugging scenarios of algorithm application elements.
- the algorithm application element is an algorithm application element used to determine whether a specific event has occurred in a business scenario.
- generate or debug an algorithm application element used to determine whether traffic congestion occurs in a road scene generate or debug an algorithm application element used to determine whether someone smokes in a public place, or generate or debug an algorithm application element used to determine whether a family scene Algorithms for whether someone fell in the meta and so on.
- the specific selection is made according to the actual situation, and the embodiment of the present disclosure makes no limitation.
- the target initial template is an initial template used to define a target image processing function to be realized in a current business scenario among at least one preset initial template.
- at least one initial template is a modularized general processing flow template corresponding to at least one image processing function.
- image processing function is an image/video analysis function, such as identifying and processing the video, it is analyzed whether the vehicle in the video has a traffic accident, or whether the person in the video has fallen, etc.
- the corresponding general processing flow may include: detection and analysis of images or video frame images to obtain detection results; detection results may include specific targets contained in images or video frame images, such as pedestrians or vehicles; attribute detection results Extraction, such as image feature extraction, obtains attribute extraction results, and encapsulates detection results and attribute extraction results as event outputs.
- its corresponding modularized general processing flow template can include: a detection module for the detection and analysis process, a network model for image/video analysis, and attribute extraction corresponding to the attribute extraction process module, and a template for the definition of the processing object type, the definition of the detection target, and the definition information of the attribute category name to be extracted in the general processing flow.
- the target image processing function can include image recognition of congestion, traffic accidents, etc.; Image recognition of someone having an accident, etc.
- each modular processing flow in the initial template does not include the actual function implementation, but only the definition of the function module, the incoming and outgoing parameter information, and the execution process description of the general processing flow, such as the function module A description of the calling relationship between and so on.
- the template configuration item may correspond to a preset target requirement, and is used to define configuration item information of at least one functional module to be implemented in the general processing flow template.
- template configuration items can be set by a user such as a template developer based on the specified target initial template and according to preset target requirements, and an application meta-template creation instruction can be generated according to the template configuration item and the template path of the target initial template , sent to the electronic device.
- a management service may run on the electronic device; wherein the management service is used to receive and manage at least one initial template, and the management service may be deployed on the electronic device, or may be deployed in a wired or wireless connection with the electronic device.
- the specific selection is made according to the actual situation, which is not limited in the embodiments of the present disclosure.
- the electronic device may receive and archive at least one initial template by running the management service, and generate at least one template path corresponding to the at least one initial template.
- a user such as a template developer can specify a target initial template for generating a target algorithm application meta-template from at least one initial template by setting a template path.
- at least one initial template may be uploaded to the management service in batches from other services in the form of a template package, or may be uploaded separately, which is not specifically limited in this embodiment of the present disclosure.
- an algorithm bin service may run on the electronic device; wherein, the algorithm bin service is used to generate and manage algorithm application elements.
- the algorithm warehouse service can receive the call of the application meta-template creation instruction through the preset creation interface, and execute the program or code that generates the algorithm application meta-template implemented in the preset creation interface.
- the application meta-template creation instruction may be sent by other devices to the algorithm warehouse service on the electronic device.
- the electronic device may be a background server, and the client device sends the application meta-template creation instruction to the algorithm warehouse service on the background server; the application meta-template creation instruction may also be sent to the algorithm warehouse service by other services on the electronic device , the specific selection is made according to the actual situation, which is not limited by the embodiments of the present disclosure.
- the electronic device can use the algorithm warehouse service to implement S101 by executing the process of S1011-S1013, as shown in FIG. 3 , as follows:
- the electronic device may receive an application meta-template creation instruction from the preset creation interface through the running algorithm store service.
- the electronic device can obtain the template configuration item and the target template path from the application meta-template creation instruction by executing the program or code of the generation algorithm application meta-template implemented in the preset creation interface; here, the target template path It is the path information corresponding to the target initial template in the management service.
- the application meta-template creation instruction includes a specified template path, and the electronic device can determine and obtain the target initial template from at least one initial template according to the template path.
- the target initial template is a modularized general processing flow template that realizes the target image processing function in a business scenario.
- the target initial template may be a newly created template that does not include historical template configuration items, or may be a historical algorithm application meta-template that includes historical template configuration items, and the electronic device may create instructions based on the currently received application meta-template.
- the historical template configuration items in the target initial template are updated to generate a new target algorithm application meta-template, which is selected according to the actual situation, and is not limited in this embodiment of the present disclosure.
- the electronic device can update or write the content of the corresponding process part in the target initial template according to the template configuration item, thereby generating
- the target algorithm applies a meta-template.
- the target initial template may include an initial process configuration template and an initial attribute configuration template; here, the initial process configuration template is used to define the general processing flow of the target image processing function, and the initial attribute configuration template is used to define the target initial template. attribute information.
- the electronic device can update or write its corresponding content items in the initial process configuration template or the initial attribute configuration template to obtain the target algorithm application meta-template. That is to say, applying the meta-template by the target algorithm also updates the target initial template of the initial process configuration template and the initial attribute configuration template.
- the initial process configuration template and the initial attribute configuration template can be templates in the form of independent files, or other data forms, such as data tables included in the target initial template, etc., according to actual conditions The selection is not limited by the embodiments of the present disclosure.
- the template configuration item includes: a template function configuration item and a template attribute configuration item; wherein, the template function configuration item is used to define module information of at least one functional module; the template attribute configuration item is used to configure the target algorithm to be generated Attribute information of the application meta template.
- the template attribute configuration item at least includes: at least one of template name and template version, and the electronic device can set the name, version, author, and applicable scenario for the generated target algorithm application meta-template through the template attribute configuration item.
- Attribute information such as type, software and hardware platform information, instruction information, etc.
- each functional module of at least one functional module may correspond to at least one function realization module, and the function realization module may be a pre-implemented and compiled software unit, which may be implemented by calling or loading, etc.
- the code execution method directly realizes the corresponding function.
- the module information may be information corresponding to the function realization module, such as module name, function type or software version and other information.
- the module information here does not include the specific implementation information of the function realization module, such as the acquisition address or call interface of the function module.
- the electronic device can specify which functions or configured modules are used to realize the modularized general processing flow in the target initial template, and realize the Definition of at least one functional module to be implemented in the general processing flow template.
- a function module with the same definition may correspond to at least one function realization module, for example, a function module with the same name may correspond to a function realization module realized in at least one way or at least one software version.
- the electronic device can preliminarily define the functional modules used to realize the general processing flow, so as to obtain the target algorithm application meta-template.
- further designated target function realization modules for each defined function module are obtained correspondingly.
- At least one function realization module may be pre-stored in the electronic device, or may be stored in other external storage spaces accessible by the electronic device. The specific selection is made according to the actual situation, and the embodiment of the present disclosure makes no limitation.
- the electronic device can implement S102 by performing the process of S1021-S1022, as shown in FIG. 4, as follows:
- the electronic device can update the initial function configuration items included in the initial process configuration template, such as the definition of function modules, to the ones described in the template function configuration items according to the template function configuration items and the preset template syntax.
- the module information that is, the information of the function realization module, obtains the target process configuration template.
- the target process configuration template is based on the modularized general processing process implemented by the target initial template, and further defines the corresponding configuration information of the functional modules contained in it, but does not contain the specific implementation data of the functional modules, and needs to be added
- the corresponding data processing functions, such as image processing functions, can be realized only after the data is specifically realized.
- the template function configuration item at least includes: at least one of target algorithm application model information and algorithm application type; wherein, the target algorithm application model information may be an algorithm model for realizing the target image processing function, such as a computer-based The neural network model of vision technology; the algorithm application type is used to represent the processing object type applicable to the target image processing function, and may include at least one of video type and image type.
- target algorithm application model information may be an algorithm model for realizing the target image processing function, such as a computer-based The neural network model of vision technology
- the algorithm application type is used to represent the processing object type applicable to the target image processing function, and may include at least one of video type and image type.
- the target algorithm application model can be a deep learning model trained using images or videos in business scenarios, for example, a convolutional neural network (Convolutional Neural Network, CNN) trained using images or videos in business scenarios ) model, or a shallow machine model trained using images or videos of business scenarios, for example, a Support Vector Machine (Support Vector Machine, SVM)
- a convolutional neural network Convolutional Neural Network, CNN
- a shallow machine model trained using images or videos of business scenarios
- SVM Support Vector Machine
- the electronic device can update the initial attribute configuration items included in the initial attribute configuration template according to the template attribute configuration items to obtain the target attribute configuration template, and according to the updated target process configuration template and target attribute configuration template, Get the target algorithm application meta-template.
- the preset template syntax may be golang template syntax
- the electronic device may parse out the template path and template configuration information from the application meta-template creation instruction calling the preset creation interface through the preset creation interface; here, Template configuration information can include template name, algorithm application type, template version and template default model definition.
- the electronic device may obtain the target initial template from the management service according to the template path.
- the target initial template may include a pipeline configuration file as an initial process configuration template, and a spec file as an initial property configuration template.
- the electronic device defines the algorithm application type and template default model as template function configuration items, writes the algorithm application type and template default model into the initial function configuration item in the pipeline configuration file with golang template syntax, and uses the template name and template version as templates
- the attribute configuration item updates the initial attribute configuration item in the spec file, such as the initial definition of the template name and version, so as to obtain the corresponding target algorithm application meta-template.
- the algorithm application meta-template is the realization of the function to be completed, such as the process framework for the realization of the algorithm model to be completed or the realization of the function module to be completed, that is, the algorithm application meta-template is an empty algorithm application element, which needs to be realized by
- the functional modules and generating algorithm application elements can analyze the image or video of the business scene.
- the target algorithm application unit can be used to intelligently process the image or video of the business scene to realize the target image processing function, such as a software unit for determining whether a specific event occurs in the business scene, which is the ultimate goal of the embodiments of the present disclosure .
- the electronic device can also maintain and manage the generated at least one algorithm application element template through the algorithm warehouse service, for example, perform operations such as traversal search or deletion processing on the at least one algorithm application element.
- the electronic device after the electronic device generates the target algorithm application meta-template, it can archive the target algorithm application meta-template through the algorithm warehouse service, for example, save it in a preset database.
- the same type of algorithm application meta-requirements can reuse the same algorithm application meta-template to generate corresponding target algorithm application meta-templates based on the implementation configuration information of their respective requirements, so as to improve the reusability of target algorithm application meta-templates .
- the electronic device can assign a corresponding template identifier to the generated algorithm application meta-template through the algorithm warehouse service, so that the algorithm application meta-template can be called according to the template identifier.
- the template identifier may be a template ID, or other types of identification information, which is selected according to actual conditions, and is not limited in this embodiment of the present disclosure.
- the algorithm warehouse service can pre-implement a preset generation interface for generating algorithm application elements according to the algorithm application element template.
- the algorithm warehouse service receives the generation of the target algorithm application meta template through the preset generation interface.
- Instructions, such as the generation instruction corresponding to the template ID of the target algorithm application meta-template, the implementation configuration information can be parsed from the generation instruction, and according to the implementation configuration information, at least one function implementation module corresponding to each function module can be obtained.
- the target function realization module specified for each function module so as to obtain at least one preset target function realization module.
- the implementation configuration information may be an address or link of an algorithm model or a functional component specified by a developer for a business scenario, or a call interface function of a pre-implemented function, and the like.
- the electronic device can obtain the algorithm application model or functional component according to the resolved address or link, or realize the corresponding function call by calling the interface function.
- the target function realization module is a function module that has been pre-realized.
- the implementation configuration information includes: process configuration information; wherein, the process configuration information represents call information corresponding to a pre-implemented function implementation module.
- the electronic device can obtain the corresponding function realization module through the process configuration information and call or load it into the algorithm application meta-template.
- the implementation configuration information includes: at least one of the plug-in configuration information and the model path; based on the above-mentioned FIG. 2-FIG. 4, as shown in FIG. Realized as follows:
- the electronic device may obtain a target algorithm application model from at least one pre-implemented algorithm application model according to the model path in the process configuration information, as at least one target function realization module.
- the pre-implemented at least one algorithm application model can be a trained neural network model, or a neural network model to be trained, and the subsequent training and model effect improvement will be carried out in the process of iteration and debugging of the target algorithm application element, specifically
- the selection is made according to the actual situation, and is not limited by the embodiments of the present disclosure.
- the electronic device can parse out the plug-in configuration information from the generation instruction, and obtain the target function plug-in from at least one pre-implemented function plug-in according to the plug-in configuration information as at least one target function realization module.
- the target function plug-in is a plug-in module that pre-completes the function implementation defined by the template function configuration item.
- the target plug-in can be a software development kit (Software Development Kit, SDK) packaged according to the pre-trained functional model, which is selected according to the actual situation, and is not limited in the embodiments of the present disclosure.
- the plug-in configuration information may be the call path information of the functional plug-in; it may also be information such as link information, or name, version, etc., which are selected according to actual conditions, and are not limited in this embodiment of the present disclosure.
- the target function plug-in may include: at least one of an image detection plug-in and an image attribute extraction plug-in.
- S1031 and S1032 are parallel method processes, and in actual application, one or more of them may be selected for execution according to the actual situation, and specifically selected according to the actual situation. Not limited.
- the process configuration information may also include invocation information of other types of function realization modules, such as invocation information of pre-implemented script files, etc., which are selected according to actual conditions, and are not limited in this embodiment of the present disclosure.
- the process configuration information may further include: at least one of the detection target type of the image detection plug-in, and the attribute information to be extracted corresponding to the image attribute extraction plug-in.
- the image detection plug-in may be a functional plug-in including a neural network model for multi-target detection, and the detection target type may be used to specify a detection target among at least one preset detection target object that the neural network model for multi-target detection can detect.
- the image attribute extraction plug-in can be a pre-implemented SDK for extracting at least one image attribute information from the detection result.
- the attribute information to be extracted can be used to extract at least one attribute information from the attribute extraction plug-in. Specify One or more attribute information.
- the electronic device can further instantiate and customize the target algorithm application element by configuring the detection target type corresponding to the image detection plug-in and the attribute information to be extracted corresponding to the image attribute extraction plug-in, thereby improving the generation of the target algorithm application. Yuan flexibility.
- the implementation configuration information further includes: attribute configuration information; here, the attribute configuration information is used to configure the attribute information of the algorithm application element to be generated.
- the attribute configuration information at least includes: at least one of an application meta name, an application meta version, an application meta description, and an algorithm scenario type.
- the attribute configuration information may also include author, algorithm scene type label, hardware platform information, authorization information, etc., which are selected according to actual conditions, and are not limited in this embodiment of the present disclosure.
- S104 Generate a target algorithm application element based on at least one target function realization module and a target algorithm application element template.
- the electronic device can render the template function configuration items in the target process configuration template in the target algorithm application meta-template according to at least one acquired target function realization module, and render each configuration content in the template function configuration items item, rendered as the realization data in the corresponding target function realization module, so as to complete the realization of each function module defined by the target algorithm application element template, and obtain the target algorithm application element capable of analyzing the image or video of the business scene.
- S1041 may be implemented through S1041-S1042, which will be described in conjunction with each step.
- the at least one target function realization module acquired by the electronic device may be a callable function interface of a prepackaged target function realization module.
- the electronic device can call at least one functional interface corresponding to at least one target function realization module, and obtain at least one implementation data returned by the at least one functional interface to the electronic device when at least one functional interface is called successfully.
- at least one piece of implementation data is used to actually realize at least one function module defined in the target algorithm application meta-template.
- at least one implementation data may be executable data.
- the electronic device can use each of at least one realization data to render the corresponding content configuration items in the module function configuration items, so as to update the target process configuration template and complete the adjustment of each function configuration item in the template.
- the function realization of the function module defined by the content configuration item When the function realization of each function module defined in the target algorithm application element template is completed, the target algorithm application element is obtained.
- the electronic device may also update the template attribute configuration items in the target attribute configuration module in the target algorithm application meta-template according to the attribute configuration information.
- the electronic device abstracts the general processing flow for intelligently processing images or videos of business scenarios into an initial template, and according to the preset target requirements of the actual business, initializes the target through template configuration items.
- the modularized general processing flow in the template is customized and configured to obtain the target algorithm application meta-template corresponding to the actual business requirements; the target algorithm application meta-template is further configured to realize the configuration information, and the target for realizing the actual business demand can be generated Algorithmic application meta.
- the redundant and repeated work of independently developing algorithm application elements each time according to different business requirements is reduced, thereby reducing the generation time of algorithm application elements and improving the generation efficiency of algorithm application elements.
- the algorithm application element template or the implementation configuration information can be modified according to the field data, so as to realize the rapid update and iteration of the algorithm application element, thereby improving the update efficiency of the algorithm application element.
- the electronic device after the electronic device generates the target algorithm application element, it can use the target algorithm application element to process the target image function.
- the electronic device can use the detection plug-in in the target algorithm application element to detect the service Perform target detection on the image or video in the scene to obtain the target detection result; use the attribute extraction plug-in to perform attribute extraction on the target detection result according to the attribute information to be extracted, and obtain the attribute extraction result; call the algorithm function model based on the target detection result and the attribute extraction result , get the processing result, and the processing result is used to determine whether a specific event occurs in the business scenario.
- the electronic device can further debug the obtained target algorithm application element according to the on-site data generated in the business scenario, to optimize the processing result of the algorithm application element, so as to meet the corresponding requirements of the business scenario and ensure that Perform accurate analysis on images or videos of business scenarios.
- the electronic device can update at least one of the target algorithm application element template and the target algorithm application element according to the image or video in the business scene to obtain the updated algorithm application element; and use the updated algorithm application element to update the image or target algorithm application element in the business scene
- the video is processed, and the updated processing result is obtained.
- the electronic device can use the rapidly generated algorithm application element to perform image processing, and can further update the algorithm application element on-site according to the image processing result, thereby improving the application convenience of the algorithm application element and update efficiency.
- the development client uploads the template package to the management service of the background server.
- the development client may be a development device connected to the background server; the template package is equivalent to at least one initial template.
- the management service may be a ModelManager service.
- the background server generates a template path according to the template package through the management service.
- the background server receives the template package uploaded by the development client through the preset service interface of the management service, and generates a corresponding template path for each initial template included in the template package. Its execution process is consistent with the description of the corresponding process in S101, and will not be repeated here.
- the development client sends an application meta template creation instruction to the algorithm warehouse service in the background server by calling the creation template interface of the algorithm warehouse service, and specifies the template name, algorithm application type, template version, and template path in the application meta template creation instruction , template default model.
- the background server also runs the algorithm warehouse service, and the creation template interface of the algorithm warehouse service is equivalent to the preset creation interface in the above embodiment.
- the execution process of S203 is consistent with the above-mentioned process description of S1011, and will not be repeated here. .
- the algorithm warehouse service on the background server generates a target template and obtains a template ID of the target template according to the application meta-template creation instruction.
- the algorithm warehouse service when the algorithm warehouse service receives the application meta template creation instruction through the preset creation interface, according to the template path specified in the application meta template creation instruction, obtain the target initial template from at least one initial template maintained by the management service .
- the target initial template can contain pipeline configuration files and spec configuration files.
- the algorithm warehouse service updates the pipeline configuration file in the target initial template with the preset golang template syntax according to the algorithm application type, template path, and template default model in the application meta-template creation instruction; and, according to the application meta-template creation instruction
- the target template name and template version in update the spec configuration file in the target initial template to obtain the target template.
- the target template is equivalent to the target algorithm application meta-template.
- the background server after the background server generates the target template, it can store the target template and the corresponding template ID in the preset database, so that the template in the preset database can be called later by the template ID to generate an algorithm application element.
- the development client sends a generation instruction to the algorithm warehouse server in the background server by calling the generation template interface of the algorithm warehouse service, specifying and carrying the template ID, the name of the algorithm application element, the algorithm application type, version, and description in the generation instruction , author, algorithm scene type label, model path, detection plug-in, attribute plug-in, algorithm application meta-type, detection target type, and the name of the attribute type to be filtered are used as parameters for calling the generated template interface.
- the template ID is used to specify the target template, the name of the algorithm application element, the algorithm application type, version, description, author, and algorithm scene type tags belong to the above attribute configuration information, model path, detection plug-in, attribute extraction plug-in, algorithm application Metatype, detection target type, and attribute type name to be filtered belong to the above-mentioned process configuration information.
- the generating template interface is equivalent to the preset generating interface.
- the algorithm warehouse service on the background server generates the target algorithm application element according to the generation instruction.
- the algorithm warehouse service on the background server receives the generation instruction through the generation template interface, it determines and obtains the target template from the preset database according to the template ID passed in by the generation instruction in the generation template interface, And according to the model path specified by the generation command, download the corresponding pre-trained target neural network model; according to the detection plug-in and attribute extraction plug-in specified by the generation command, determine the pre-packaged detection plug-in corresponding to the detection plug-in from the preset implementation module library SDK, and the attribute extraction SDK corresponding to the attribute extraction plug-in.
- the algorithm warehouse service can obtain the callable functional interfaces corresponding to the detection SDK and attribute extraction SDK.
- the algorithm warehouse service can render and update according to the target neural network model and the template default model defined in the pipeline configuration file of the target template; and, according to the detection target type specified in the generation instruction, call the corresponding functional interface of the detection SDK; And according to the type name of the attribute to be filtered specified in the generation instruction; call the function interface corresponding to the attribute extraction SDK.
- the function interfaces of the detection SDK and the attribute extraction SDK are called successfully, the corresponding implementation data of the detection SDK and the attribute extraction SDK are obtained.
- the algorithm warehouse service uses the corresponding implementation data of the detection SDK and attribute extraction SDK, as well as the algorithm application metatype in the generated command, to render and update the pipeline configuration file in the target template.
- the algorithm warehouse service also re-renders and updates the spec configuration file of the target template according to the attribute configuration information such as the name of the algorithm application element specified in the generation instruction, the algorithm application type, version, description, author, and algorithm scene type label , and finally get the updated pipeline configuration file and the updated spec configuration file, so as to complete the rendering of the target template and obtain the target algorithm application element.
- the attribute configuration information such as the name of the algorithm application element specified in the generation instruction, the algorithm application type, version, description, author, and algorithm scene type label , and finally get the updated pipeline configuration file and the updated spec configuration file, so as to complete the rendering of the target template and obtain the target algorithm application element.
- the algorithm warehouse service may also save the target algorithm application element and assign a corresponding application element identifier for subsequent invocation of the target algorithm application element.
- the target algorithm application element can be used to realize the target image processing function in business scenarios.
- the electronic device can generate an algorithm application meta-template according to the target initial template and template configuration information, and then generate a target algorithm application meta-template according to the implementation configuration information and the target algorithm application meta-template, so that developers do not need to manually generate algorithm applications completely by writing codes elements, and manually debug the generated algorithm application elements, which improves the generation and update efficiency of algorithm application elements.
- FIG. 8 is a schematic structural diagram of the algorithm application element generation device provided by an embodiment of the present disclosure; as shown in FIG. 8 , the algorithm application element generation device 1 includes:
- the obtaining part 11 is configured to obtain template configuration items and target initial templates by receiving application meta-template creation instructions;
- the target initial templates are modularized general processing flow templates that realize target image processing functions in business scenarios;
- the template configuration item is used to define at least one functional module to be implemented in the general processing flow template according to preset target requirements;
- the template creation part 12 is configured to generate a target algorithm application meta-template based on the template configuration item and the target initial template;
- the generating part 13 is configured to obtain at least one preset target function realization module according to the realization configuration information in the generation instruction when receiving the generation instruction for the target algorithm application meta-template; the realization The configuration information is the implementation information corresponding to the at least one function module; and the target algorithm application element is generated based on the at least one target function realization module and the target algorithm application element template.
- the acquisition part 11 is further configured to receive the application meta-template creation instruction through a preset creation interface; from the application meta-template creation instruction, parse and obtain the template configuration item and target Template path; the target template path is path information corresponding to the target initial template in the management service; according to the template path, the target initial template is obtained.
- the template configuration items include: template function configuration items and template attribute configuration items;
- the target initial template includes: an initial process configuration template and an initial attribute configuration template;
- the template creation part is further configured to According to the template function configuration item, update the initial process configuration template with the preset template syntax to obtain the target process configuration template;
- the template function configuration item includes the configuration information of the function realization module; according to the template attribute
- the configuration item is to update the initial attribute configuration template to obtain the target attribute configuration template, thereby obtaining the target algorithm application meta-template;
- the template attribute configuration item is used to configure the attribute information of the target algorithm application meta-template.
- the template function configuration items include at least:
- At least one of target algorithm application model information and algorithm application type includes at least one of video type and image type;
- the template attribute configuration items at least include:
- At least one of template name and template version At least one of template name and template version.
- the implementation configuration information includes: process configuration information; the process configuration information includes: at least one of plug-in configuration information and a model path; the generating part is further configured to, according to the model path, Determine the target algorithm application model as the at least one target function realization module; and/or, according to the plug-in configuration information, determine the pre-realized target function plug-in as the at least one target function realization module; the target function plug-in is used for Realize the functions defined by the template function configuration items.
- the target function plug-in includes: at least one of an image detection plug-in and an image attribute extraction plug-in; the process configuration information further includes:
- the algorithm application element generating device further includes a receiving module configured to obtain at least one preset target function realization module according to the realization configuration information in the generation instruction , receiving a generation instruction for applying a meta-template for the target algorithm through a preset generation interface.
- the generating part 13 is further configured to render the template function configuration items in the target process configuration template according to the at least one target function realization module, so as to obtain the target algorithm application element .
- the generation part 13 is further configured to call at least one function interface corresponding to the at least one target function realization module, and if the call of the at least one function interface is successful, obtain the At least one implementation data returned by at least one functional interface; using the at least one implementation data to render the template function configuration item to obtain the target algorithm application element.
- the implementation configuration information further includes: attribute configuration information; the generating part is further configured to update template attribute configuration items in the target attribute configuration template according to the attribute configuration information; the attribute configuration The information is used to configure the attributes of the algorithm application element; the attribute configuration information at least includes: at least one of the application element name, the application element version, the application element description, and the algorithm scene type.
- the target algorithm application meta-template is stored in a preset database; the algorithm application meta-generating device further includes a management part configured to respond to the management of the target algorithm application meta-template An instruction to search or delete the target algorithm application meta-template by traversing the algorithm application meta-template in the preset database.
- FIG. 9 is a schematic structural diagram of the electronic device provided by the embodiment of the present disclosure.
- the electronic device 2 includes: a memory 22 and a processor 23, wherein the memory 22 and the processing The device 23 is connected through a communication bus 24; the memory 22 is used to store executable algorithm application element generation instructions; the processor 23 is used to implement the method provided by the embodiment of the present disclosure when executing the executable computer program stored in the memory 22, for example , the algorithm application element generation method provided by the embodiment of the present disclosure.
- the embodiment of the present disclosure provides a computer-readable storage medium, which stores an executable algorithm application element generation instruction, which is used to cause the processor 23 to implement the method provided by the embodiment of the present disclosure, for example, the algorithm provided by the embodiment of the present disclosure Applied meta generation method.
- An embodiment of the present disclosure provides a computer program product, including a computer program or an instruction.
- the computer program or instruction is executed by a processor, the above method for generating an algorithm application element is implemented.
- the computer readable storage is transferred to a computer readable storage medium which may be volatile or nonvolatile.
- the computer-readable storage medium may be memory such as FRAM, ROM, PROM, EPROM, EEPROM, flash memory, magnetic surface memory, optical disk, or CD-ROM; it may also be various devices including one or any combination of the above memories.
- executable algorithm application element generation instructions may be in the form of programs, software, software modules, scripts or codes, in any form of programming language (including compiled or interpreted language, or declarative or procedural language) and can be deployed in any form, including as a stand-alone program or as a module, component, subroutine, or other unit suitable for use in a computing environment.
- executable algorithm application meta-generated instructions may, but do not necessarily correspond to files in a file system, may be stored as part of a file that holds other programs or data, for example, in a hypertext markup language (HTML, Hyper Text Markup Language) document, in a single file dedicated to the program in question, or in multiple cooperating files (for example, in a file that stores one or more modules, subroutines, or code sections file).
- HTML Hyper Text Markup Language
- the executable algorithm application element generation instructions can be deployed to be executed on one computing device, or on multiple computing devices located at one site, or, alternatively, at multiple sites and interconnected by a communication network. Execution on multiple computing devices.
- the algorithm application element generation method, device, electronic equipment, computer program product, and computer-readable storage medium abstract the general processing flow for intelligently processing images or videos of actual business scenarios into target initial templates, and According to the preset target requirements of the actual business, the template configuration items are used to customize the modularized general processing flow in the target initial template to obtain the target algorithm application meta-template corresponding to the actual business needs; for the target algorithm application meta-template, further By configuring and implementing the configuration information, the target algorithm application element used to realize the actual business requirements can be generated. In this way, the redundant and repeated work of independently developing algorithm application elements each time according to different business requirements is reduced, thereby reducing the generation time of algorithm application elements and improving the generation efficiency of algorithm application elements. Moreover, when debugging at the business site, the algorithm application element template or the implementation configuration information can be modified according to the field data, so as to realize the rapid update and iteration of the algorithm application element, thereby improving the update efficiency of the algorithm application element.
Landscapes
- Engineering & Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Theoretical Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Stored Programmes (AREA)
Abstract
本公开实施例公开了一种算法应用元生成方法、装置、电子设备、计算机程序产品及计算机可读存储介质,能够提高算法应用元的生成与更新效率。该方法包括:通过接收应用元模板创建指令,获取模板配置项与目标初始模板;目标初始模板为在业务场景中,实现目标图像处理功能的模块化的通用处理流程模板;模板配置项用于根据预设目标需求,在通用处理流程模板中定义待实现的至少一个功能模块;基于模板配置项与目标初始模板,生成目标算法应用元模板;在接收到针对目标算法应用元模板的生成指令的情况下,根据生成指令中的实现配置信息,获取预设的至少一个目标功能实现模块;基于至少一个目标功能实现模块与目标算法应用元模板,生成目标算法应用元。
Description
相关申请的交叉引用
本公开基于申请号为202111232488.7、申请日为2021年10月22日、申请名称为“算法应用元生成方法、装置、设备及计算机可读存储介质”的中国专利申请提出,并要求该中国专利申请的优先权,该中国专利申请的全部内容在此以引入方式并入本公开。
本公开涉及计算机视觉技术,尤其涉及一种算法应用元生成方法、装置、电子设备、计算机程序产品及计算机可读存储介质。
算法应用元是用于对图像或视频进行智能分析的软件单元,能够应用于各种业务场景,例如,人流量预警场景、交通事故路段预警场景等。算法应用元的生成过程包括了创建算法应用元和调试算法应用元。
相关技术中,创建算法应用元和调试算法应用元,都需要由开发人员手动进行,例如,每次在创建一个新的算法应用元时,都需要重新进行一次独立开发过程,因此产生了大量重复的算法应用元开发工作,导致开发时间和人力成本呈线性增长。并且,在实际业务场景中调试算法应用元时,由于安全性与隐私性要求,业务场景的现场数据难以回流至算法应用元的开发场景中进行调试,从而加长了算法应用元的更新迭代周期。由此可见,相关技术中,需要较长时间才能生成与更新算法应用元,从而使得算法应用元的生成效率与更新效率较低。
发明内容
本公开实施例至少提供一种算法应用元生成方法、装置、电子设备、计算机程序产品及计算机可读存储介质,能够提高算法应用元的生成与更新效率。
本公开实施例的技术方案是这样实现的:
本公开实施例提供一种算法应用元生成方法,包括:
通过接收应用元模板创建指令,获取模板配置项与目标初始模板;所述目标初始模板为在业务场景中,实现目标图像处理功能的模块化的通用处理流程模板;所述模板配置项用于根据预设目标需求,在所述通用处理流程模板中定义待实现的至少一个功能模块;
基于所述模板配置项与所述目标初始模板,生成目标算法应用元模板;
在接收到针对所述目标算法应用元模板的生成指令的情况下,根据所述生成指令中的实现配置信息,获取预设的至少一个目标功能实现模块;所述实现配置信息为所述至少一个功能模块对应的实现信息;
基于所述至少一个目标功能实现模块与所述目标算法应用元模板,生成目标算法应用元。
本公开实施例提供一种算法应用元生成装置,包括:
获取部分,被配置为通过接收应用元模板创建指令,获取模板配置项与目标初始模板;所述目标初始模板为在业务场景中,实现目标图像处理功能的模块化的通用处理流程模板; 所述模板配置项用于根据预设目标需求,在所述通用处理流程模板中定义待实现的至少一个功能模块;
模板创建部分,被配置为基于所述模板配置项与所述目标初始模板,生成目标算法应用元模板;
生成部分,被配置为在接收到针对所述目标算法应用元模板的生成指令的情况下,根据所述生成指令中的实现配置信息,获取预设的至少一个目标功能实现模块;所述实现配置信息为所述至少一个功能模块对应的实现信息;并基于所述至少一个目标功能实现模块与所述目标算法应用元模板,生成目标算法应用元。
本公开实施例提供一种电子设备,包括:
存储器,被配置为存储可执行算法应用元生成指令;
处理器,被配置为执行所述存储器中存储的可执行算法应用元生成指令时,实现本公开实施例提供的算法应用元生成方法。
本公开实施例提供一种计算机可读存储介质,存储有可执行算法应用元生成指令,被配置为引起处理器执行时,实现本公开实施例提供的算法应用元生成方法。
本公开实施例提供一种计算机程序产品,包括计算机程序或指令,所述计算机程序或指令被处理器执行时,实现上述的算法应用元生成方法。
本公开实施例具有如下技术效果:将针对实际业务场景的图像或视频进行智能处理的通用处理流程抽象为目标初始模板,并根据实际业务的预设目标需求,通过模板配置项对目标初始模板中的模块化的通用处理流程进行自定义配置,得到实际业务需求对应的目标算法应用元模板;针对目标算法应用元模板,进一步配置实现配置信息,即可生成用于实现实际业务需求的目标算法应用元。这样,减少了每次根据不同业务需求独立开发算法应用元的冗余重复工作,从而减少了算法应用元的生成时间,提高了算法应用元的生成效率。并且,在业务现场进行调试时,可以根据现场数据,对算法应用元模板或者实现配置信息进行修改,即可实现对算法应用元的快速更新迭代,从而提高了算法应用元的更新效率。
应当理解的是,以上的一般描述和后文的细节描述仅是示例性和解释性的,而非限制本公开。
此处的附图被并入说明书中并构成本说明书的一部分,这些附图示出了符合本公开的实施例,并与说明书一起用于说明本公开的技术方案。
图1是本公开实施例提供的算法应用元生成***100的一个可选的架构示意图;
图2是本公开实施例提供的算法应用元生成方法的一种流程示意图一;
图3是本公开实施例提供的算法应用元生成方法的可选的流程示意图二;
图4是本公开实施例提供的算法应用元生成方法的可选的流程示意图三;
图5是本公开实施例提供的算法应用元生成方法的可选的流程示意图四;
图6是本公开实施例提供的算法应用元生成方法的可选的流程示意图五;
图7是本公开实施例提供的在实际应用场景中生成算法应用元的流程示意图;
图8为本公开实施例提供的算法应用元生成装置的结构示意图;
图9为本公开实施例提供的电子设备的结构示意图。
为了使本公开的目的、技术方案和优点更加清楚,下面将结合附图对本公开作进一步地详细描述,所描述的实施例不应视为对本公开的限制,本领域普通技术人员在没有做出创造 性劳动前提下所获得的所有其它实施例,都属于本公开保护的范围。
在以下的描述中,涉及到“一些实施例”,其描述了所有可能实施例的子集,但是可以理解,“一些实施例”可以是所有可能实施例的相同子集或不同子集,并且可以在不冲突的情况下相互结合。
对本公开实施例进行进一步详细说明之前,对本公开实施例中涉及的名词和术语进行说明,本公开实施例中涉及的名词和术语适用于如下的解释。
1)人工智能(Artificial Intelligence,AI)是利用数字计算机或者数字计算机控制的机器模拟、延伸和扩展人的智能,感知环境、获取知识并使用知识获得最佳结果的理论、方法、技术及应用***。换句话说,人工智能是计算机科学的一个综合技术,它企图了解智能的实质,并生成一种新的能以人类智能相似的方式做出反应的智能机器。人工智能也就是研究各种智能机器的设计原理与实现方法,使机器具有感知、推理与决策的功能。
人工智能技术是一门综合学科,设计领域广泛,既有硬件层面的技术也有软件层面的技术。人工智能基础技术一般包括如传感器、专用人工智能芯片、云计算、分布式存储、大数据处理技术、操作/交互***、机电一体化等技术。人工智能软件技术包括计算机视觉技术、语音处理技术、自然语言处理技术以及机器学习/深度学习等几大方向。
2)计算机视觉技术(Computer Vision,CV)计算机实际是一门研究如何使机器“看”的科学,更进一步的说,就是指用摄影机和电脑代替人眼对目标进行识别、跟踪和测量等机器视觉,并进一步做图像处理,使电脑处理成为更适合人眼观察或传送给仪器检测的图像。作为一个科学学科,计算机视觉研究相关的理论和技术,试图建立能够从图像或者多维数据中获取信息的人工智能***。计算机视觉技术通常包括图像处理、图像识别、图像语义理解、图像检索、OCR、视频分析、视频语义理解、视频内容/行为识别、三维物体重建、3D技术、虚拟现实、增强现实、同步定位与地图构建等技术,还包括常见的人脸识别、指纹识别等生物特征识别技术。
3)图像/视频分析,是指对图像采集设备所采集到的图像进行分析处理,从而明确图像或视频中是否出现了特定事件。例如,通过对视频进行识别处理,分析出视频中的车辆是否出现了交通事故,或者是视频中的人是否跌倒等。
4)算法应用元,是用于对图像或视频进行分析处理的软件模块,其可以集成图像或视频分析时所需要的算法模型,以及调度服务组件、消息服务组件和存储服务组件等。将指定场景下的图像或视频输入到算法应用元中,就可以得到该场景对应的分析结果。在一些实施例中,还可以继续基于算法应用元进行开发,例如,在算法应用元的基础上继续开发界面交互功能,通信功能等,从而基于算法应用元得到一个完整的应用软件。
算法应用元是用于对图像或视频进行智能分析的软件单元,能够应用于各种业务场景,例如,人流量预警场景、交通事故路段预警场景等。算法应用元的生成过程包括了创建算法应用元和调试算法应用元。
目前,算法应用元种类非常繁多,应用场景也非常广泛,尤其是来自安防和智慧城市行业的大量定制化开发需求,对算法应用元开发的敏捷性越来越高。相关技术中的算法应用元开发,是每发布一个算法模型之后,都重新开发一次算法应用元。示例性地,相关技术在创建一个新的算法应用元时,需要手动编写算法应用元的各个文件,例如算法应用元的目录结构、代码文件、配置文件、文档文件和脚本文件(Makefile等)的内容,以及手动编写在代码文件中手动编写代码等,从而使得生成算法应用元开发时间和人力成本呈线性增长,降低了算法应用元的生成效率。
并且,对于算法应用元在应用现场的迭代问题,很多应用现场由于自身安全性的考虑,不允许现场数据的回流到算法应用元的开发组织处进行训练迭代,难以根据现场数据及时对算法应用元进行迭代更新,导致算法应用元的迭代周期十分漫长。并且,即便应用现场允许数据回流,根据目前算法应用元的迭代过程,也要重新进行模型训练与算法更新,迭代周期 较长,从而大大降低了算法应用元的更新效率。
由上述可见,相关技术中,算法应用元的生成和更新都需要较长时间,无法快速生成算法应用元并进行更新迭代,从而降低了算法应用元的生成与更新效率。
本公开实施例提供一种算法应用元生成方法、装置、电子设备、计算机程序产品及计算机可读存储介质,能够提高算法应用元生成与更新效率。下面说明本公开实施例提供的电子设备的示例性应用,本公开实施例提供的电子设备可以实施为笔记本电脑,平板电脑,台式计算机,机顶盒,移动设备(例如,移动电话,便携式音乐播放器,个人数字助理,专用消息设备,便携式游戏设备)等各种类型的用户终端,也可以实施为服务器。下面,将说明电子设备实施为服务器的示例性应用。
参见图1,图1是本公开实施例提供的算法应用元生成***100的一个可选的架构示意图。终端400通过网络300连接服务器200,网络300可以是广域网或者是局域网,又扩展是二者的组合。
其中,终端400归属于业务场景,如项目调试现场的工程开发人员,开发人员可以使用终端400,根据实际的业务需求向服务器200发送应用元模板创建指令;
服务器200可以是后台服务器,用于通过接收应用元模板创建指令,获取模板配置项与目标初始模板;目标初始模板为在业务场景中,实现目标图像处理功能的模块化的通用处理流程模板;模板配置项用于根据预设目标需求,在通用处理流程模板中定义待实现的至少一个功能模块;基于模板配置项与目标初始模板,生成目标算法应用元模板;在接收到针对目标算法应用元模板的生成指令的情况下,根据生成指令中的实现配置信息,获取预设的至少一个目标功能实现模块;实现配置信息为至少一个功能模块对应的实现信息;基于至少一个目标功能实现模块与目标算法应用元模板,生成目标算法应用元。
服务器200,还用于将生成的目标算法应用元发送给终端400,以使终端400通过执行目标算法应用元,对业务场景中的视频或图像进行分析处理,得到处理结果;以及,根据处理结果对目标算法应用元进行进一步的配置更新等操作。
一些实施例中,服务器200可以是独立的物理服务器,也可以是多个物理服务器构成的服务器集群或者分布式***,还可以是提供云服务、云数据库、云计算、云函数、云存储、网络服务、云通信、中间件服务、域名服务、安全服务、CDN、以及大数据和人工智能平台等基础云计算服务的云服务器。终端400可以是智能手机、平板电脑、笔记本电脑、台式计算机、智能音箱、智能手表等,但并不局限于此。终端以及服务器可以通过有线或无线通信方式进行直接或间接地连接,本公开实施例中不做限制。
下面,将结合本公开实施例提供的电子设备的示例性应用和实施,说明本公开实施例提供的算法应用元生成方法。
参见图2,图2是本公开实施例提供的算法应用元生成方法的可选的流程示意图一,将结合图2示出的步骤进行说明。
S101、通过接收应用元模板创建指令,获取模板配置项与目标初始模板;目标初始模板为在业务场景中,实现目标图像处理功能的模块化的通用处理流程模板;模板配置项用于根据预设目标需求,在通用处理流程模板中定义待实现的至少一个功能模块。
本公开实施例适用于算法应用元的生成与调试场景。这里,算法应用元用于判定业务场景中是否发生了特定事件的算法应用元。示例性地,生成或调试用于判定道路场景中是否发生了交通拥堵实现的算法应用元,生成或调试用于判定公共场合是否有人吸烟的算法应用元,或者,生成或调试用于判定家庭场景中是否有人摔倒的算法应用元等等。具体的根据实际情况进行选择,本公开实施例不作限定。
本公开实施例中,目标初始模板为预设的至少一个初始模板中,用于定义当前的业务场景下待实现的目标图像处理功能的初始模板。其中,至少一个初始模板为至少一种图像处理功能对应的模块化的通用处理流程模板。示例性地,在上述的图像处理功能为图像/视频分析 功能,如通过对视频进行识别处理,分析出视频中的车辆是否出现了交通事故,或者是视频中的人是否跌倒等功能的情况下,相应的通用处理流程可以包括:对图像或视频帧图像进行检测分析,得到检测结果;检测结果可以包括图像或视频帧图像中包含的特定目标,如行人或车辆等等;对检测结果进行属性提取,如图像特征提取,得到属性提取结果,将检测结果和属性提取结果封装为事件输出。可以看出,对于上述的通用处理流程,其对应的模块化的通用处理流程模板可以包含:用于检测分析过程的检测模块、用于图像/视频分析的网络模型、属性提取过程对应的属性提取模块,以及通用处理流程的处理对象类型定义、检测目标定义、以及待提取的属性种类名定义信息的模板。
需要说明的是,不同的业务场景可以对应不同的目标图像处理功能。例如,当业务场景为城市主干道时,目标图像处理功能可以包括对拥堵、车祸等现象的图像识别;当业务场景为家庭内部时,目标图像处理功能可以包括对是否有不法人员闯入、是否有人发生意外等的图像识别等等。
需要说明的是,初始模板中的各个模块化的处理流程并不包含实际的功能实现,仅仅为功能模块的定义、传入传出的参数信息、以及通用处理流程的执行过程描述,如功能模块之间调用关系的描述等等。
本公开实施例中,模板配置项可以是预设目标需求对应的,用于在通用处理流程模板中定义待实现的至少一个功能模块的配置项信息。在一些实施例中,模板配置项可由用户如模板开发人员,基于指定的目标初始模板,根据预设目标需求来设定,并根据模板配置项与目标初始模板的模板路径生成应用元模板创建指令,发送给电子设备。
在一些实施例中,电子设备上可以运行有管理服务;其中,管理服务用于接收并管理至少一个初始模板,管理服务可以部署在电子设备上,也可以部署在与电子设备以有线或无线的方式连接的其他设备上,具体的根据实际情况进行选择,本公开实施例不作限定。
在一些实施例中,对于管理服务部署在电子设备上的情况,电子设备可以通过运行管理服务,接收至少一个初始模板并存档,以及,生成至少一个初始模板对应的至少一个模板路径。这样,用户如模板开发人员可以通过设置模板路径,从至少一个初始模板中指定用于生成目标算法应用元模板的目标初始模板。这里,至少一个初始模板可以以模板包的形式,由其他服务批量上传至管理服务,也可以分别单独上传,本公开实施例不作具体的限定。
在一些实施例中,电子设备上可以运行有算法仓服务;其中,算法仓服务用于生成与管理算法应用元。算法仓服务可以通过预设创建接口,接收应用元模板创建指令的调用,并执行预设创建接口中实现的生成算法应用元模板的程序或代码。这里,应用元模板创建指令可以是其他设备发送至电子设备上的算法仓服务的。示例性地,电子设备可以是后台服务器,客户端设备将应用元模板创建指令发送至后台服务器上的算法仓服务;应用元模板创建指令也可以是电子设备上的其他服务发送给算法仓服务的,具体的根据实际情况进行选择,本公开实施例不作限定。
在一些实施例中,电子设备可以利用算法仓服务,通过执行S1011-S1013的过程实现S101,如图3所示,如下:
S1011、通过预设创建接口,接收应用元模板创建指令。
本公开实施例中,电子设备可以通过运行的算法仓服务,从预设创建接口接收到应用元模板创建指令。
S1012、从应用元模板创建指令中,解析得到模板配置项与目标模板路径。
本公开实施例中,电子设备可以通过执行预设创建接口中实现的生成算法应用元模板的程序或代码,从应用元模板创建指令中解析得到模板配置项与目标模板路径;这里,目标模板路径为目标初始模板在管理服务中对应的路径信息。
S1013、根据模板路径,得到目标初始模板。
本公开实施例中,应用元模板创建指令中包含指定的模板路径,电子设备可以根据模板 路径,从至少一个初始模板中确定并获取到目标初始模板。这里,目标初始模板为在业务场景中,实现目标图像处理功能的模块化的通用处理流程模板。
进一步的,目标初始模板可以是新建的、不包含历史模板配置项的模板,也可以是包含历史模板配置项的历史算法应用元模板,电子设备可以根据当前接收到的应用元模板创建指令,对目标初始模板中的历史模板配置项进行更新,以生成新的目标算法应用元模板,具体的根据实际情况进行选择,本公开实施例不作限定。
S102、基于模板配置项与目标初始模板,生成目标算法应用元模板。
本公开实施例中,由于模板配置项是针对于目标初始模板中通用处理流程配置信息,电子设备可以根据模板配置项,对目标初始模板中对应的流程部分的内容进行更新或写入,从而生成目标算法应用元模板。
在一些实施例中,目标初始模板可以包含初始流程配置模板与初始属性配置模板;这里,初始流程配置模板用于定义目标图像处理功能的通用处理流程,初始属性配置模板用于定义目标初始模板的属性信息。电子设备可以根据接收到的模板配置项,对其在初始流程配置模板或初始属性配置模板中各自对应的内容项进行更新或写入,得到目标算法应用元模板。也就是说,目标算法应用元模板也即更新了初始流程配置模板与初始属性配置模板的目标初始模板。
在一些实施例中,初始流程配置模板与初始属性配置模板可以是独立的文件形式的模板,也可以是其他数据形式,如包含在目标初始模板中的数据表等等,具体的根据实际情况进行选择,本公开实施例不作限定。
在一些实施例中,模板配置项包括:模板功能配置项与模板属性配置项;其中,模板功能配置项用于定义至少一个功能模块的模块信息;模板属性配置项用于配置所要生成的目标算法应用元模板的属性信息。
在一些实施例中,模板属性配置项至少包括:模板名称与模板版本中的至少一个,电子设备可以通过模板属性配置项,为所生成的目标算法应用元模板设置名称、版本、作者、适用场景类型、软硬件平台信息、使用说明信息等等属性信息。
需要说明的是,本公开实施例中,至少一个功能模块的每个功能模块可以对应至少一种功能实现模块,功能实现模块可以是预先实现的、经过编译的软件单元,可通过调用或加载等代码执行方式,直接实现相应功能。模块信息可以是功能实现模块对应的信息,如模块名称、功能类型或软件版本等信息。这里模块信息并不包含功能实现模块具体的实现信息,如功能模块的获取地址或调用接口等等。电子设备可以根据模板功能配置项中包含的模块信息,在目标初始模板中待实现的至少一个功能模块中,指定使用哪些功能或配置的模块实现目标初始模板中模块化的通用处理流程,实现对通用处理流程模板中待实现的至少一个功能模块的定义。
本公开实施例中,同一定义的功能模块可对应至少一个的功能实现模块,如相同名称的功能模块可对应以至少一种方式实现或至少一种软件版本对应的功能实现模块。电子设备通过模板配置项,可以先初步定义出用于实现通用处理流程的功能模块,从而得到目标算法应用元模板。在接收到针述目标算法应用元模板的生成指令的情况下,再根据生成指令中的实现配置信息,相应地获取为已定义的各个功能模块的进一步指定的目标功能实现模块。
在一些实施例中,至少一个功能实现模块可以预先存储在电子设备中,也可以存储在电子设备可访问的其他外部存储空间。具体的根据实际情况进行选择,本公开实施例不作限定。
在一些实施例中,基于图3,电子设备可以通过执行S1021-S1022的过程实现S102,如图4所示,如下:
S1021、根据模板功能配置项,以预设模板语法更新初始流程配置模板,得到目标流程配置模板。
本公开实施例中,电子设备可以根据模板功能配置项,按照预设模板语法,将初始流程 配置模板中包含的初始功能配置项,如对功能模块的定义,更新为模板功能配置项中描述的模块信息,也即功能实现模块的信息,得到目标流程配置模板。
需要说明的是,目标流程配置模板是在目标初始模板实现的模块化的通用处理流程基础上,进一步定义了其中包含的功能模块对应配置信息,但并不包含功能模块具体的实现数据,需要添加具体的实现数据之后,才能实现相应的数据处理功能,如图像处理功能。
在一些实施例中,模板功能配置项至少包括:目标算法应用模型信息与算法应用类型中的至少一个;其中,目标算法应用模型信息可以是用于实现目标图像处理功能的算法模型,如基于计算机视觉技术的神经网络模型;算法应用类型用于表征目标图像处理功能适用的处理对象类型,可以包含视频类型与图像类型中的至少一个。
在一些实施例中,目标算法应用模型可以是利用业务场景下的图像或视频训练出的深度学习模型,例如,利用业务场景下的图像或视频训练出的卷积神经网络(Convolutional Neural Network,CNN)模型,也可以是利用业务场景的图像或视频训练出的浅层机器模型,例如,支持向量机(Support Vector Machine,SVM)本公开在此不作限定。
S1022、根据模板属性配置项,更新初始属性配置模板,得到目标属性配置模板,从而得到目标算法应用元模板。
本公开实施例中,电子设备可以根据模板属性配置项,对于初始属性配置模板包含的初始属性配置项进行更新,得到目标属性配置模板,并根据更新得到的目标流程配置模板与目标属性配置模板,得到目标算法应用元模板。
在一些实施例中,预设模板语法可以是golang模板语法,电子设备可以通过预设创建接口,从调用预设创建接口的应用元模板创建指令中,解析出模板路径与模板配置信息;这里,模板配置信息可以包含模板名称、算法应用类型、模板版本与模板默认模型定义。电子设备可以根据模板路径,从管理服务中获取到目标初始模板,示例性地,目标初始模板中可以包含pipeline配置文件作为初始流程配置模板,以及包含spec文件作为初始属性配置模板。电子设备将算法应用类型与模板默认模型定义作为模板功能配置项,以golang模板语法将算法应用类型与模板默认模型写入pipeline配置文件中的初始功能配置项,并将模板名称与模板版本作为模板属性配置项,对spec文件中的初始属性配置项,如对模板名称与版本的初始定义进行更新,从而得到对应的目标算法应用元模板。
可以理解的是,算法应用元模板为待完成功能实现,如待完成算法模型的实现或待完成功能模块的实现的流程框架,也即算法应用元模板是空的算法应用元,需要通过实现其中的功能模块,生成算法应用元,才能够对业务场景的图像或视频进行分析。
需要说明的是,目标算法应用元可以是用于对业务场景的图像或视频进行智能处理,实现目标图像处理功能,如确定业务场景是否发生特定事件的软件单元,是本公开实施例的最终目标。
在一些实施例中,电子设备还可以通过算法仓服务维护和管理生成的至少一个算法应用元模板,示例性地,对至少一个算法应用元进行遍历查找或删除处理等操作。
S103、在接收到针对目标算法应用元模板的生成指令的情况下,根据生成指令中的实现配置信息,获取预设的至少一个目标功能实现模块;实现配置信息为至少一个功能模块对应的实现信息。
本公开实施例中,电子设备在生成目标算法应用元模板之后,可以通过算法仓服务,将目标算法应用元模板进行存档,如保存在预设数据库中。这样,在实际应用场景中,同类型的算法应用元需求可以复用相同的算法应用元模板以各自需求的实现配置信息生成相应目标算法应用元,以提高目标算法应用元模板的可复用性。
本公开实施例中,电子设备可以通过算法仓服务,对生成的算法应用元模板分配相应的模板标识,以使算法应用元模板可以根据模板标识进行调用。在一些实施例中,模板标识可以是模板ID,也可以是其他类型的标识信息,具体的根据实际情况进行选择,本公开实施例 不作限定。
在一些实施例中,算法仓服务中可以预先实现有用于根据算法应用元模板生成算法应用元的预设生成接口,算法仓服务在通过预设生成接口,接收到针对目标算法应用元模板的生成指令,如目标算法应用元模板的模板ID对应的生成指令的情况下,可以从生成指令中解析出实现配置信息,根据实现配置信息,从每个功能模块对应的至少一个功能实现模块中,获取为每个功能模块指定的目标功能实现模块,从而获取到预设的至少一个目标功能实现模块。示例性地,实现配置信息可以是开发人员为业务场景所指定的算法模型或功能组件的地址或链接,或预实现功能的调用接口函数等等。电子设备可以依据解析出的地址或链接获取到算法应用模型或功能组件,或通过调用接口函数实现相应的功能调用。这里,目标功能实现模块是已经预实现完成的功能模块。
在一些实施例中,实现配置信息包括:流程配置信息;其中,流程配置信息表征预实现的功能实现模块对应的调用信息。电子设备可以通过流程配置信息,获取对应的功能实现模块并调用或加载至算法应用元模板。
在一些实施例中,实现配置信息包括:插件配置信息与模型路径中的至少一个;基于上述图2-图4,如图5所示,S103可以通过执行S1031与S1032中的至少一个流程步骤来实现,如下:
S1031、根据模型路径,确定目标算法应用模型为至少一个目标功能实现模块。
本公开实施例中,电子设备可以根据流程配置信息中的模型路径,从预实现的至少一个算法应用模型中获取目标算法应用模型,作为至少一个目标功能实现模块。
这里,预实现的至少一个算法应用模型可以是训练好的神经网络模型,也可以是待训练的神经网络模型,后续在目标算法应用元的迭代与调试的过程中进行训练与模型效果提升,具体的根据实际情况进行选择,本公开实施例不作限定。
S1032、根据插件配置信息,确定目标功能插件为至少一个目标功能实现模块。
本公开实施例中,电子设备可以从生成指令中解析出插件配置信息,根据插件配置信息,从预先实现的至少一个功能插件中获取目标功能插件,作为至少一个目标功能实现模块。这里,目标功能插件为预先完成了模板功能配置项所定义的功能实现的插件模块。目标插件可以是根据预训练的功能模型封装得到的软件开发工具包(Software Development Kit,SDK),具体的根据实际情况进行选择,本公开实施例不作限定。
在一些实施例中,插件配置信息可以是至少一个,相应地,目标功能插件也可以是至少一个。
在一些实施例中,插件配置信息可以是功能插件的调用路径信息;也可以是链接信息、或者名称、版本等信息,具体的根据实际情况进行选择,本公开实施例不作限定。
在一些实施例中,在目标图像处理功能为单帧检测图像处理功能的情况下,目标功能插件可以包括:图像检测插件、图像属性提取插件中的至少一个。
需要说明的是,本公开实施例中,S1031与S1032为并列的方法流程,实际应用中可以根据实际情况,选择其中的一个或多个来执行,具体的根据实际情况进行选择,本公开实施例不作限定。
在一些实施例中,流程配置信息也可以包含其他类型的功能实现模块的调用信息,如预实现的脚本文件的调用信息等等,具体的根据实际情况进行选择,本公开实施例不作限定。
在一些实施例中,流程配置信息还可以包括:图像检测插件的检测目标类型,与图像属性提取插件对应的待提取属性信息中的至少一个。这里,图像检测插件可以是包含多目标检测的神经网络模型的功能插件,检测目标类型可以用于在多目标检测的神经网络模型所能检测的预设的至少一种检测目标对象中,指定一种或多种检测目标对象的类型。图像属性提取插件可以是预实现的,用于从检测结果中提取出至少一种图像属性信息的SDK,待提取属性信息可以用于从属性提取插件所能提取的至少一种属性信息中,指定一种或多种属性信息。
可以理解的是,电子设备可以通过对图像检测插件对应的检测目标类型与图像属性提取插件对应的待提取属性信息的配置,对目标算法应用元进行进一步的实例化定制,从而提高生成目标算法应用元的灵活性。
在一些实施例中,实现配置信息还包括:属性配置信息;这里,属性配置信息用于配置所要生成的算法应用元的属性信息。在一些实施例中,属性配置信息至少包括:应用元名称、应用元版本、应用元描述、算法场景类型中的至少一个。或者,属性配置信息还可以包括作者、算法场景类型标签、硬件平台信息、授权信息等等,具体的根据实际情况进行选择,本公开实施例不作限定。
S104、基于至少一个目标功能实现模块与目标算法应用元模板,生成目标算法应用元。
本公开实施例中,电子设备可以根据获取到的至少一个目标功能实现模块,对目标算法应用元模板中,目标流程配置模板中的模板功能配置项进行渲染,将模板功能配置项中各个配置内容项,渲染为相应的目标功能实现模块中的实现数据,从而完成对目标算法应用元模板定义的各个功能模块的实现,得到能够实现对业务场景的图像或视频进行分析的目标算法应用元。
在一些实施例中,基于图4,如图6所示,S1041可以通过S1041-S1042来实现,将结合各步骤进行说明。
S1041、对至少一个目标功能实现模块对应的至少一个功能接口进行调用,在至少一个功能接口调用成功的情况下,得到至少一个功能接口返回的至少一个实现数据。
在一些实施例中,电子设备所获取到的至少一个目标功能实现模块可以是预封装的目标功能实现模块的可调用的功能接口。电子设备可以对至少一个目标功能实现模块对应的至少一个功能接口进行调用,在至少一个功能接口均调用成功的情况下,得到至少一个功能接口向电子设备返回的至少一个实现数据。其中,至少一个实现数据用于真正实现目标算法应用元模板中定义的至少一个功能模块。在一些实施例中,至少一个实现数据可以是可执行数据。
S1042、利用至少一个实现数据,对模板功能配置项进行渲染,得到目标算法应用元。
本公开实施例中,电子设备可以使用至少一个实现数据中的每个实现数据,对模块功能配置项中相应的内容配置项进行渲染,从而更新目标流程配置模板,完成对模板功能配置项中各个内容配置项所定义的功能模块的功能实现。在完成对目标算法应用元模板中所定义的每个功能模块的功能实现的情况下,得到目标算法应用元。
在一些实施例中,在实现配置信息中还包括属性配置信息的情况下,电子设备还可以根据属性配置信息,更新目标算法应用元模板中,目标属性配置模块中的模板属性配置项。
可以理解的是,本公开实施例中,电子设备将针对业务场景的图像或视频进行智能处理的通用处理流程抽象为初始模板,并根据实际业务的预设目标需求,通过模板配置项对目标初始模板中的模块化的通用处理流程进行自定义配置,得到实际业务需求对应的目标算法应用元模板;针对目标算法应用元模板,进一步配置实现配置信息,即可生成用于实现实际业务需求的目标算法应用元。这样,减少了每次根据不同业务需求独立开发算法应用元的冗余重复工作,从而减少了算法应用元的生成时间,提高了算法应用元的生成效率。并且,在业务现场进行调试时,可以根据现场数据,对算法应用元模板或者实现配置信息进行修改,即可实现对算法应用元的快速更新迭代,从而提高了算法应用元的更新效率。
在一些实施例中,电子设备在生成目标算法应用元之后,就可以使用目标算法应用元进行目标图像功能的处理了,电子设备可以通过目标算法应用元中的检测插件,基于检测目标类型对业务场景中的图像或视频进行目标检测,得到目标检测结果;利用属性提取插件,根据待提取属性信息对目标检测结果进行属性提取,得到属性提取结果;基于目标检测结果与属性提取结果调用算法功能模型,得到处理结果,处理结果用于确定业务场景中是否发生特定事件。
在一些实施例中,电子设备可以根据业务场景中产生的现场数据,对得到的目标算法应 用元进行进一步的调试,来优化算法应用元的处理结果,以满足业务场景的对应的需求,保证能够对业务场景的图像或视频进行准确的分析。
电子设备可以根据业务场景中的图像或视频,对目标算法应用元模板与目标算法应用元中的至少一个进行更新,得到更新算法应用元;并利用更新算法应用元,对业务场景中的图像或视频进行处理,得到更新处理结果。
可以理解的是,本公开实施例中,电子设备可以使用快速生成的算法应用元进行图像处理,并且还可以根据图像处理结果进一步对算法应用元进行现场更新,从而提高了算法应用元的应用便捷性和更新效率。
下面,将基于图7,以电子设备实施为后台服务器为例。对本公开实施例在实际应用场景中的实现过程进行介绍。
S201、开发客户端将模板包上传至后台服务器的管理服务。
S201中,开发客户端可以是与后台服务器连接的开发设备;模板包相当于至少一个初始模板。管理服务可以是ModelManager服务。
S202、后台服务器通过管理服务,根据模板包生成模板路径。
S202中,后台服务器通过管理服务的预设服务接口,接收开发客户端上传的模板包,并为模板包中包含的各个初始模板生成对应的模板路径。其执行过程与S101中的对应过程描述一致,此处不再赘述。
S203、开发客户端通过调用算法仓服务的创建模板接口,向后台服务器中的算法仓服务发送应用元模板创建指令,在应用元模板创建指令中指定模板名称、算法应用类型、模板版本、模板路径、模板默认模型。
S203中,后台服务器上还运行有算法仓服务,算法仓服务的创建模板接口相当于上述实施例中的预设创建接口,S203的执行过程与上述的S1011的过程描述一致,此处不再赘述。
S204、后台服务器上的算法仓服务根据应用元模板创建指令,生成目标模板并得到目标模板的模板ID。
S204中,算法仓服务在通过预设创建接口,接收到应用元模板创建指令的情况下,根据应用元模板创建指令中指定的模板路径,从管理服务维护的至少一个初始模板中获取目标初始模板。这里,目标初始模板可以包含pipeline配置文件与spec配置文件。算法仓服务根据应用元模板创建指令中的算法应用类型、模板路径与模板默认模型,以预设的golang模板语法,对目标初始模板中的pipeline配置文件进行更新;并且,根据应用元模板创建指令中的目标模板名称与模板版本,对目标初始模板中的spec配置文件进行更新,得到目标模板。这里,目标模板相当于目标算法应用元模板。
这里,后台服务器生成目标模板之后,可以将目标模板与对应的模板ID保存在预设数据库中,以便后续通过模板ID对预设数据库中的模板进行调用,生成算法应用元。
S205、开发客户端通过调用算法仓服务的生成模板接口,向后台服务器中的算法仓服务器发送生成指令,在生成指令中指定并携带模板ID、算法应用元的名称、算法应用类型、版本、描述、作者、算法场景类型标签、模型路径、检测插件、属性插件、算法应用元类型、检测目标类型、需过滤的属性种类名,作为生成模板接口的调用传入参数。
S205中,模板ID用于指定目标模板,算法应用元的名称、算法应用类型、版本、描述、作者、算法场景类型标签属于上述的属性配置信息,模型路径、检测插件、属性提取插件、算法应用元类型、检测目标类型、需过滤的属性种类名属于上述的流程配置信息。这里,生成模板接口相当于预设生成接口。
S206、后台服务器上的算法仓服务根据生成指令生成目标算法应用元。
S206中,后台服务器上的算法仓服务在通过生成模板接口,接收到生成指令的情况下,根据生成指令在生成模板接口中传入的模板ID,从预设数据库中确定并获取到目标模板,并根据生成指令指定的模型路径,下载得到对应的预训练的目标神经网络模型;根据生成指令 指定的检测插件与属性提取插件,从预设的实现模块库中确定检测插件对应的预封装的检测SDK,以及属性提取插件对应的属性提取SDK。并且,算法仓服务可以获取检测SDK与属性提取SDK各自对应的可调用的功能接口。
S206中,算法仓服务可以根据目标神经网络模型,目标模板的pipeline配置文件中定义的模板默认模型进行渲染与更新;并且,根据生成指令中指定的检测目标类型,调用检测SDK对应的功能接口;以及根据生成指令中指定的需过滤的属性种类名;调用属性提取SDK对应的功能接口。在检测SDK与属性提取SDK的功能接口调用成功的情况下,得到检测SDK与属性提取SDK各自对应的实现数据。算法仓服务进而使用检测SDK与属性提取SDK各自对应的实现数据,以及生成指令中的算法应用元类型,对目标模板中的pipeline配置文件进行渲染与更新。
S206中,算法仓服务还根据生成指令中指定的算法应用元的名称、算法应用类型、版本、描述、作者以及算法场景类型标签等属性配置信息,对目标模板的spec配置文件进行重新渲染与更新,最终得到更新的pipeline配置文件与更新的spec配置文件,从而完成对目标模板的渲染,得到目标算法应用元。
在一些实施例中,算法仓服务生成目标算法应用元之后,还可以将目标算法应用元进行保存并分配相应的应用元标识,以供后续对目标算法应用元进行调用。目标算法应用元可以用于在业务场景中实现目标图像处理功能。
由上述可知,电子设备可以根据目标初始模板与模板配置信息生成算法应用元模板,然后根据实现配置信息与目标算法应用元模板生成目标算法应用元,从而无需开发人员完全通过编写代码手动生成算法应用元,以及手动对生成的算法应用元进行调试,提高了算法应用元的生成与更新效率。
本公开还提供一种算法应用元生成装置,图8为本公开实施例提供的算法应用元生成装置的结构示意图;如图8所示,算法应用元生成装置1包括:
获取部分11,被配置为通过接收应用元模板创建指令,获取模板配置项与目标初始模板;所述目标初始模板为在业务场景中,实现目标图像处理功能的模块化的通用处理流程模板;所述模板配置项用于根据预设目标需求,在所述通用处理流程模板中定义待实现的至少一个功能模块;
模板创建部分12,被配置为基于所述模板配置项与所述目标初始模板,生成目标算法应用元模板;
生成部分13,被配置为在接收到针对所述目标算法应用元模板的生成指令的情况下,根据所述生成指令中的实现配置信息,获取预设的至少一个目标功能实现模块;所述实现配置信息为所述至少一个功能模块对应的实现信息;并基于所述至少一个目标功能实现模块与所述目标算法应用元模板,生成目标算法应用元。
在一些实施例中,所述获取部分11,还被配置为通过预设创建接口,接收所述应用元模板创建指令;从所述应用元模板创建指令中,解析得到所述模板配置项与目标模板路径;所述目标模板路径为所述目标初始模板在管理服务中对应的路径信息;根据所述模板路径,得到所述目标初始模板。
在一些实施例中,所述模板配置项包括:模板功能配置项与模板属性配置项;所述目标初始模板包括:初始流程配置模板与初始属性配置模板;所述模板创建部分,还被配置为根据所述模板功能配置项,以所述预设模板语法更新所述初始流程配置模板,得到目标流程配置模板;所述模板功能配置项包含所述功能实现模块的配置信息;根据所述模板属性配置项,更新所述初始属性配置模板,得到目标属性配置模板,从而得到所述目标算法应用元模板;所述模板属性配置项用于配置所述目标算法应用元模板的属性信息。
在一些实施例中,所述模板功能配置项至少包括:
目标算法应用模型信息与算法应用类型中的至少一个;其中,所述算法应用类型包含视 频类型与图像类型中的至少一个;
所述模板属性配置项至少包括:
模板名称与模板版本中的至少一个。
在一些实施例中,所述实现配置信息包括:流程配置信息;所述流程配置信息包括:插件配置信息与模型路径中的至少一个;所述生成部分,还被配置为根据所述模型路径,确定目标算法应用模型为所述至少一个目标功能实现模块;和/或,根据所述插件配置信息,确定预先实现的目标功能插件为所述至少一个目标功能实现模块;所述目标功能插件用于实现所述模板功能配置项所定义的功能。
在一些实施例中,所述目标功能插件包括:图像检测插件、图像属性提取插件中的至少一个;所述流程配置信息还包括:
所述图像检测插件对应的检测目标类型,与所述图像属性提取插件对应的待提取属性信息中的至少一个。
在一些实施例中,所述算法应用元生成装置还包括接收模块,所述接收模块,被配置为所述根据所述生成指令中的实现配置信息,获取预设的至少一个目标功能实现模块之前,通过预设生成接口,接收针对所述目标算法应用元模板的生成指令。
在一些实施例中,所述生成部分13,还被配置为根据所述至少一个目标功能实现模块,对所述目标流程配置模板中的模板功能配置项进行渲染,从而得到所述目标算法应用元。
在一些实施例中,所述生成部分13,还被配置为对所述至少一个目标功能实现模块对应的至少一个功能接口进行调用,在所述至少一个功能接口调用成功的情况下,得到所述至少一个功能接口返回的至少一个实现数据;利用所述至少一个实现数据,对所述模板功能配置项进行渲染,得到所述目标算法应用元。
在一些实施例中,所述实现配置信息还包括:属性配置信息;所述生成部分,还被配置为根据所述属性配置信息,更新目标属性配置模板中的模板属性配置项;所述属性配置信息用于配置所述算法应用元的属性;所述属性配置信息至少包括:应用元名称、应用元版本、应用元描述、算法场景类型中的至少一个。
在一些实施例中,所述目标算法应用元模板保存在预设数据库中;所述算法应用元生成装置还包括管理部分,所述管理部分,被配置为响应于针对目标算法应用元模板的管理指令,通过遍历所述预设数据库中的算法应用元模板,对所述目标算法应用元模板进行查找或删除。
需要说明的是,以上装置实施例的描述,与上述方法实施例的描述是类似的,具有同方法实施例相似的有益效果。对于本公开装置实施例中未披露的技术细节,请参照本公开方法实施例的描述而理解。
本公开实施例还提供一种电子设备,图9为本公开实施例提供的电子设备的结构示意图,如图9所示,电子设备2包括:存储器22和处理器23,其中,存储器22和处理器23通过通信总线24连接;存储器22,用于存储可执行算法应用元生成指令;处理器23,用于执行存储器22中存储的可执行计算机程序时,实现本公开实施例提供的方法,例如,本公开实施例提供的算法应用元生成方法。
本公开实施例提供一种计算机可读存储介质,存储有可执行算法应用元生成指令,用于引起处理器23执行时,实现本公开实施例提供的方法,例如,本公开实施例提供的算法应用元生成方法。
本公开实施例提供一种计算机程序产品,包括计算机程序或指令,所述计算机程序或指令被处理器执行时,实现上述的算法应用元生成方法。
在本公开的一些实施例中,计算机可读存储转接至可为易失性或非易失的计算机可读取存储介质。计算机可读存储介质可以是FRAM、ROM、PROM、EPROM、EEPROM、闪存、磁表面存储器、光盘、或CD-ROM等存储器;也可以是包括上述存储器之一或任意组合的各种设备。
在本公开的一些实施例中,可执行算法应用元生成指令可以采用程序、软件、软件模块、脚本或代码的形式,按任意形式的编程语言(包括编译或解释语言,或者声明性或过程性语言)来编写,并且其可按任意形式部署,包括被部署为独立的程序或者被部署为模块、组件、子例程或者适合在计算环境中使用的其它单元。
作为示例,可执行算法应用元生成指令可以但不一定对应于文件***中的文件,可以可被存储在保存其它程序或数据的文件的一部分,例如,存储在超文本标记语言(HTML,Hyper Text Markup Language)文档中的一个或多个脚本中,存储在专用于所讨论的程序的单个文件中,或者,存储在多个协同文件(例如,存储一个或多个模块、子程序或代码部分的文件)中。
作为示例,可执行算法应用元生成指令可被部署为在一个计算设备上执行,或者在位于一个地点的多个计算设备上执行,又或者,在分布在多个地点且通过通信网络互连的多个计算设备上执行。
以上所述,仅为本公开的实施例而已,并非用于限定本公开的保护范围。凡在本公开的精神和范围之内所作的任何修改、等同替换和改进等,均包含在本公开的保护范围之内。
本公开实施例提供的算法应用元生成方法、装置、电子设备、计算机程序产品及计算机可读存储介质,将针对实际业务场景的图像或视频进行智能处理的通用处理流程抽象为目标初始模板,并根据实际业务的预设目标需求,通过模板配置项对目标初始模板中的模块化的通用处理流程进行自定义配置,得到实际业务需求对应的目标算法应用元模板;针对目标算法应用元模板,进一步配置实现配置信息,即可生成用于实现实际业务需求的目标算法应用元。这样,减少了每次根据不同业务需求独立开发算法应用元的冗余重复工作,从而减少了算法应用元的生成时间,提高了算法应用元的生成效率。并且,在业务现场进行调试时,可以根据现场数据,对算法应用元模板或者实现配置信息进行修改,即可实现对算法应用元的快速更新迭代,从而提高了算法应用元的更新效率。
Claims (15)
- 一种算法应用元生成方法,包括:通过接收应用元模板创建指令,获取模板配置项与目标初始模板;所述目标初始模板为在业务场景中,实现目标图像处理功能的模块化的通用处理流程模板;所述模板配置项用于根据预设目标需求,在所述通用处理流程模板中定义待实现的至少一个功能模块;基于所述模板配置项与所述目标初始模板,生成目标算法应用元模板;在接收到针对所述目标算法应用元模板的生成指令的情况下,根据所述生成指令中的实现配置信息,获取预设的至少一个目标功能实现模块;所述实现配置信息为所述至少一个功能模块对应的实现信息;基于所述至少一个目标功能实现模块与所述目标算法应用元模板,生成目标算法应用元。
- 根据权利要求1所述的方法,其中,所述通过接收应用元模板创建指令,获取模板配置项与目标初始模板,包括:通过预设创建接口,接收所述应用元模板创建指令;从所述应用元模板创建指令中,解析得到所述模板配置项与目标模板路径;所述目标模板路径为所述目标初始模板在管理服务中对应的路径信息;根据所述模板路径,得到所述目标初始模板。
- 根据权利要求1或2所述的方法,其中,所述模板配置项包括:模板功能配置项与模板属性配置项;所述目标初始模板包括:初始流程配置模板与初始属性配置模板;所述基于所述模板配置项与所述目标初始模板,生成目标算法应用元模板,包括:根据所述模板功能配置项,以所述预设模板语法更新所述初始流程配置模板,得到目标流程配置模板;所述模板功能配置项包含所述功能实现模块的配置信息;根据所述模板属性配置项,更新所述初始属性配置模板,得到目标属性配置模板,从而得到所述目标算法应用元模板;所述模板属性配置项用于配置所述目标算法应用元模板的属性信息。
- 根据权利要求3所述的方法,其中,所述模板功能配置项至少包括:目标算法应用模型信息与算法应用类型中的至少一个;其中,所述算法应用类型包含视频类型与图像类型中的至少一个;所述模板属性配置项至少包括:模板名称与模板版本中的至少一个。
- 根据权利要求1、2或4所述的方法,其中,所述实现配置信息包括:流程配置信息;所述流程配置信息包括:插件配置信息与模型路径中的至少一个;所述根据所述生成指令中的实现配置信息,获取预设的至少一个目标功能实现模块,包括:根据所述模型路径,确定目标算法应用模型为所述至少一个目标功能实现模块;和/或,根据所述插件配置信息,确定目标功能插件为所述至少一个目标功能实现模块;所述目标功能插件用于实现所述模板功能配置项所定义的功能。
- 根据权利要求5所述的方法,其中,所述目标功能插件包括:图像检测插件、图像属性提取插件中的至少一个;所述流程配置信息还包括:所述图像检测插件对应的检测目标类型,与所述图像属性提取插件对应的待提取属性信息中的至少一个。
- 根据权利要求1、2、4或6所述的方法,其中,所述根据所述生成指令中的实现配置信息,获取预设的至少一个目标功能实现模块之前,所述方法还包括:通过预设生成接口,接收针对所述目标算法应用元模板的生成指令。
- 根据权利要求3所述的方法,其中,所述基于所述至少一个目标功能实现模块与所述目标算法应用元模板,生成目标算法应用元,包括:根据所述至少一个目标功能实现模块,对所述目标流程配置模板中的模板功能配置项进行渲染,从而得到所述目标算法应用元。
- 根据权利要求8所述的方法,其中,所述根据所述至少一个目标功能实现模块,对所述目标流程配置模板中的模板功能配置项进行渲染,从而得到所述目标算法应用元,包括:对所述至少一个目标功能实现模块对应的至少一个功能接口进行调用,在所述至少一个功能接口调用成功的情况下,得到所述至少一个功能接口返回的至少一个实现数据;利用所述至少一个实现数据,对所述模板功能配置项进行渲染,得到所述目标算法应用元。
- 根据权利要求8或9所述的方法,其中,所述实现配置信息还包括:属性配置信息;所述方法还包括:根据所述属性配置信息,更新所述目标属性配置模板中的模板属性配置项;所述属性配置信息用于配置所述算法应用元的属性;所述属性配置信息至少包括:应用元名称、应用元版本、应用元描述、算法场景类型中的至少一个。
- 根据权利要求1、2、4、6、8或9所述的方法,其中,所述目标算法应用元模板保存在预设数据库中;所述方法还包括:响应于针对目标算法应用元模板的管理指令,通过遍历所述预设数据库中的算法应用元模板,对所述目标算法应用元模板进行查找或删除。
- 一种算法应用元生成装置,包括:获取部分,被配置为通过接收应用元模板创建指令,获取模板配置项与目标初始模板;所述目标初始模板为在业务场景中,实现目标图像处理功能的模块化的通用处理流程模板;所述模板配置项用于根据预设目标需求,在所述通用处理流程模板中定义待实现的至少一个功能模块;模板创建部分,被配置为基于所述模板配置项与所述目标初始模板,生成目标算法应用元模板;生成部分,被配置为在接收到针对所述目标算法应用元模板的生成指令的情况下,根据所述生成指令中的实现配置信息,获取预设的至少一个目标功能实现模块;所述实现配置信息为所述至少一个功能模块对应的实现信息;基于所述至少一个目标功能实现模块与所述目标算法应用元模板,生成目标算法应用元。
- 一种电子设备,包括:存储器,被配置为存储可执行算法应用元生成指令;处理器,被配置为执行所述存储器中存储的可执行算法应用元生成指令时,实现权利要求1至11任一项所述的方法。
- 一种计算机可读存储介质,存储有可执行算法应用元生成指令,被配置为引起处理器执行时,实现权利要求1至11任一项所述的方法。
- 一种计算机程序产品,包括计算机程序或指令,在所述计算机程序或指令被处理器执行时,实现权利要求1至11中任一项所述的方法。
Applications Claiming Priority (2)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202111232488.7A CN114005055A (zh) | 2021-10-22 | 2021-10-22 | 算法应用元生成方法、装置、设备及计算机可读存储介质 |
CN202111232488.7 | 2021-10-22 |
Publications (1)
Publication Number | Publication Date |
---|---|
WO2023065746A1 true WO2023065746A1 (zh) | 2023-04-27 |
Family
ID=79923947
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
PCT/CN2022/107174 WO2023065746A1 (zh) | 2021-10-22 | 2022-07-21 | 算法应用元生成方法、装置、电子设备、计算机程序产品及计算机可读存储介质 |
Country Status (2)
Country | Link |
---|---|
CN (1) | CN114005055A (zh) |
WO (1) | WO2023065746A1 (zh) |
Families Citing this family (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN114005055A (zh) * | 2021-10-22 | 2022-02-01 | 深圳市商汤科技有限公司 | 算法应用元生成方法、装置、设备及计算机可读存储介质 |
CN115617819B (zh) * | 2022-12-19 | 2023-03-14 | 思创数码科技股份有限公司 | 数据存储方法、***、计算机设备及存储介质 |
CN117555586B (zh) * | 2024-01-11 | 2024-03-22 | 之江实验室 | 一种算法应用发布、管理及评分方法 |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111241067A (zh) * | 2020-01-14 | 2020-06-05 | 智恒科技股份有限公司 | 一种水务大数据挖掘建模管理方法、装置、设备和介质 |
CN111767026A (zh) * | 2020-06-22 | 2020-10-13 | 北京百度网讯科技有限公司 | 业务对象模板的生成方法、装置、设备和介质 |
CN113282396A (zh) * | 2021-06-15 | 2021-08-20 | 深圳市商汤科技有限公司 | 图像处理方法、***、装置、计算机设备及存储介质 |
CN114005055A (zh) * | 2021-10-22 | 2022-02-01 | 深圳市商汤科技有限公司 | 算法应用元生成方法、装置、设备及计算机可读存储介质 |
-
2021
- 2021-10-22 CN CN202111232488.7A patent/CN114005055A/zh not_active Withdrawn
-
2022
- 2022-07-21 WO PCT/CN2022/107174 patent/WO2023065746A1/zh unknown
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111241067A (zh) * | 2020-01-14 | 2020-06-05 | 智恒科技股份有限公司 | 一种水务大数据挖掘建模管理方法、装置、设备和介质 |
CN111767026A (zh) * | 2020-06-22 | 2020-10-13 | 北京百度网讯科技有限公司 | 业务对象模板的生成方法、装置、设备和介质 |
CN113282396A (zh) * | 2021-06-15 | 2021-08-20 | 深圳市商汤科技有限公司 | 图像处理方法、***、装置、计算机设备及存储介质 |
CN114005055A (zh) * | 2021-10-22 | 2022-02-01 | 深圳市商汤科技有限公司 | 算法应用元生成方法、装置、设备及计算机可读存储介质 |
Also Published As
Publication number | Publication date |
---|---|
CN114005055A (zh) | 2022-02-01 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
WO2023065746A1 (zh) | 算法应用元生成方法、装置、电子设备、计算机程序产品及计算机可读存储介质 | |
CN110825362B (zh) | 低代码应用软件开发***及方法 | |
US20230057335A1 (en) | Deployment of self-contained decision logic | |
US10642913B2 (en) | Intent and bot based query guidance | |
CN112199086B (zh) | 自动编程控制***、方法、装置、电子设备及存储介质 | |
KR20200068050A (ko) | 인공지능 수행을 위한 학습 데이터 생성장치 및 방법 | |
CN111061475B (zh) | 软件代码生成方法、装置、计算机设备和存储介质 | |
CN111242317B (zh) | 管理应用的方法、装置、计算机设备和存储介质 | |
JP2019510282A (ja) | グロッサリを用いたストリーミング分析アプリケーションの生成 | |
CN116775183A (zh) | 基于大语言模型的任务生成方法、***、设备及存储介质 | |
US11762636B2 (en) | Method and system for implementing application lineage metadata and registration | |
CN110851488A (zh) | 基于多源多模态数据融合分析处理方法及平台 | |
WO2023040151A1 (zh) | 算法应用元生成方法、装置、电子设备、计算机可读存储介质及计算机程序产品 | |
CN108469977B (zh) | 一种界面数据管理方法 | |
CN114065948A (zh) | 构建预训练模型的方法、装置、终端设备和存储介质 | |
CN112699042B (zh) | 一种单元测试案例的生成方法及装置 | |
CN116974581B (zh) | 代码生成方法、装置、电子设备和存储介质 | |
CN113987337A (zh) | 基于组件化动态编排的搜索方法、***、设备及存储介质 | |
CN109063059A (zh) | 行为日志处理方法、装置及电子设备 | |
CN117234505A (zh) | 一种交互页面生成方法、装置、设备及其存储介质 | |
CN116383787A (zh) | 页面创建方法、装置、计算机设备及存储介质 | |
CN116226850A (zh) | 应用程序的病毒检测方法、装置、设备、介质及程序产品 | |
CN115543428A (zh) | 一种基于策略模板的模拟数据生成方法和装置 | |
CN114721930A (zh) | 一种数据处理方法、装置、设备及介质 | |
CN113645064B (zh) | 任务下发方法、装置、电子设备及存储介质 |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
121 | Ep: the epo has been informed by wipo that ep was designated in this application |
Ref document number: 22882368 Country of ref document: EP Kind code of ref document: A1 |
|
NENP | Non-entry into the national phase |
Ref country code: DE |