Intelligent highway artificial intelligence application open platform and method, medium and electronic equipment

Abstract

The invention provides an intelligent highway artificial intelligence application open platform and method, a medium and electronic equipment, and relates to the technical field of intelligent traffic. The platform comprises an expressway data set which is used for collecting actual business scene data of an expressway and uploading the data to the platform to form data set assets; the expressway model library comprises a pre-training model and various algorithm models, and model assets are formed; the expressway mirror image library comprises various algorithm operation environment packaging mirror images to form mirror image assets; the expressway algorithm library deploys various algorithms to the platform in a servitization manner to form algorithm assets; the front-end display module is used for displaying a graphical interface to a user; and the A I application service module is used for realizing a platform application service management function and comprises a cloud management sub-module and an edge management sub-module. According to the method, cloud edge collaboration can be realized, and a one-stop and full-process AI application development solution is provided.

Description

Technical Field

[0001] The present invention belongs to the field of intelligent transportation technology, and in particular relates to an open platform, method, medium and electronic equipment for intelligent highway artificial intelligence applications. Background Art

[0002] With the development of artificial intelligence, various artificial intelligence frameworks, deep learning platforms, etc. are constantly emerging. These artificial intelligence infrastructures play an important role in promoting the development of artificial intelligence.

[0003] At present, some companies have launched AI open platforms. These platforms are based on the core concept of "inclusive AI" and lowering the threshold for AI application, and their functions are similar. The construction of AI platforms based on public clouds is usually driven by commercial interests, with the characteristics of "heavy cloud and weak edge" and payment by function. Some open platforms in the field of intelligent transportation lack the ability to design, integrate and deploy AI models, and have slow integration of upstream and downstream resources. It is difficult for ordinary users to use these platforms to build their own solutions, which seriously restricts the rapid engineering implementation of intelligent transportation products. Summary of the invention

[0004] To this end, the technical problem to be solved by the present invention is to provide an open platform, method, medium and electronic equipment for smart highway artificial intelligence applications, which can realize cloud-edge collaboration and provide a one-stop, full-process AI application development solution.

[0005] In a first aspect, the present invention discloses an open platform for intelligent highway artificial intelligence applications, the platform comprising:

[0006] Highway data set, which collects actual highway business scenario data and uploads it to the platform to form data set assets for AI algorithm model training and development;

[0007] Highway model library, including pre-trained models and various algorithm models, forming model assets for AI model training acceleration or inference service deployment;

[0008] The highway image library includes packaged images of various algorithm operating environments to form image assets for deployment of training environment inference services;

[0009] Highway algorithm library, which deploys various algorithms as services to the platform to form algorithm assets and is used to provide algorithm service interfaces;

[0010] A front-end display module, connected to the highway data set, the highway model library, the highway mirror library and the highway algorithm library, for displaying a graphical interface to the user;

[0011] AI application service module, used to implement the platform application service management function; including cloud management submodule and edge management submodule;

[0012] The cloud management submodule is used to implement cloud service management, including an online service unit and an algorithm demonstration unit; the online service unit is used for user online management and online reasoning to view the returned results; the algorithm demonstration unit is used to support multiple algorithm demonstrations, upload images, videos, audio or text according to the instructions, and view the algorithm results in real time;

[0013] The edge management submodule enables users to manage edge application services of road sections, including a section management unit, an edge model management unit, an edge service management unit and an edge annotation service management unit; the section management unit is used to enable the platform to manage the road sections supported by the edge; the edge model management unit is used to implement model distribution, as well as management and deployment of edge model versions on the cloud; the edge service management unit is used to implement application distribution, as well as management and deployment of edge application services on the cloud; the edge annotation service management unit is used to implement data feedback, as well as edge annotation service details and annotation data viewing.

[0014] Furthermore, the platform also includes an AI data management module, and the AI ​​data management module includes:

[0015] The data set management submodule is used for users to operate batch data sets and get quick responses, including a data set creation unit, a data set information modification unit, a data set deletion unit, a data set query unit, and a data set publishing unit;

[0016] The annotation task management submodule is used for users to manage annotation tasks. The annotation methods include manual annotation and automatic annotation. The manual annotation is performed by users through a graphical interface to annotate data. The automatic annotation is performed by selecting an annotation service to automatically annotate the selected data set. The annotation task management submodule includes an annotation task creation unit, an annotation task dispatching unit, a data annotation unit, an annotation task result review unit, an annotation task result viewing unit, and an annotation task batch deletion unit.

[0017] The annotation service management submodule is used for users to manage annotation tasks to achieve the automatic annotation; the annotation service management submodule includes an annotation service creation unit, an annotation service start and stop unit, an annotation service editing unit and an annotation service batch deletion unit.

[0018] Furthermore, in the data set management submodule,

[0019] The data set creation unit includes uploading images, texts, and audio data to form a data set or importing an existing data set, and inputting a data set name, data type, annotation type, label type, label group, and data set description during creation;

[0020] The data set information modification unit includes modifying the name, data type, annotation type, label type, label group and data set description of the created data set;

[0021] The data set deletion unit is used to delete the created data set;

[0022] The data set query unit queries the data set according to the input number or name;

[0023] The data set publishing unit publishes a version of the selected data set, including a version number and publishing information;

[0024] In the annotation task management submodule,

[0025] The labeling task creation unit creates a labeling task by inputting a task name and selecting a data set;

[0026] The labeling task dispatching unit splits the created labeling task and dispatches it to multiple users;

[0027] The data annotation unit is used for users to annotate the data set to which the annotation task belongs, and the annotation content includes image target area and type, image segmentation, video target tracking, speech recognition, and text classification and segmentation;

[0028] The labeling task result review unit is used for users with review authority to review completed labeling tasks and re-label unqualified data;

[0029] The labeling task result viewing unit is used for users to view data labeling results;

[0030] The labeling task batch deletion unit is used to delete the labeling tasks in batches;

[0031] In the annotation service management submodule,

[0032] The annotation service creation unit is used to create an annotation service by inputting a service name and selecting a model type, an algorithm type, a model version, an image version, a node type and specification, and a service quantity in a graphical interface;

[0033] The annotation service start and stop unit is used to start and stop the selected annotation service;

[0034] The annotation service editing unit is used to modify the selected annotation service, including modifying the model type, algorithm type, model version, image version, node type and specification, and service quantity;

[0035] The annotation service batch deletion unit is used to delete the batch annotation services.

[0036] Furthermore, the platform also includes an AI computing power management module, and the AI ​​computing power management module includes:

[0037] The cluster status query unit is used to check the status of the platform cluster and monitor the health status of the platform cluster;

[0038] Resource specification management unit, used for managing platform container resources;

[0039] Node label management unit, used to manage platform computing nodes;

[0040] A resource management unit is used to manage platform computing resources, and the resource management unit includes:

[0041] The platform GPU virtualization subunit is used to divide a GPU into multiple virtual GPUs according to the quantity and provide them to different users;

[0042] The authority and quota management subunit is used for user authority management to achieve user-level management of computing power, and user quota management to achieve on-demand allocation of computing power;

[0043] The resource pool creation / deletion subunit is used by administrators to create or delete platform resource pools;

[0044] The computing power metering subunit is used to count the total amount of computing power actually used by users for billing operations.

[0045] Furthermore, the platform also includes an AI algorithm training control module, and the AI ​​algorithm training control module includes:

[0046] The image management submodule is used to manage the platform images, including image uploading, image editing and image batch deletion;

[0047] The model management submodule is used to manage the platform model weight files and optimization tasks, including:

[0048] Model weight management unit, including model weight list viewing, uploading, and version management, is used to view model weight file information, and users upload model weights to the platform and perform version management;

[0049] A model optimization task management unit, used to input a task name and a task description in the graphical interface, select an optimization type, an optimization algorithm, a model and a data set to create a model optimization task, and start, stop, modify and delete the model optimization task;

[0050] The algorithm development unit is used by algorithm development users to implement custom algorithm development. The algorithm development unit includes:

[0051] Notebook management subunit, used by users to manage Notebook algorithm running scripts;

[0052] Algorithm management subunit, used by users to implement algorithm code management;

[0053] The training task management subunit is used by users to independently create and start and stop training tasks, and to monitor resources and view logs of training tasks;

[0054] The terminal management unit is used for users to manage remote connections on the platform.

[0055] Furthermore, the platform also includes an AI mall module, which includes a mall page submodule, a product management submodule, and a service provider management submodule;

[0056] The mall page submodule is used for data algorithm product display and detail viewing, including the mall homepage, login page, and product detail page;

[0057] The product management submodule is used by administrators to manage the entire process of the mall products, including product addition, editing, listing application, listing review, and delisting;

[0058] The service provider management submodule is used to manage third-party service provider members on the platform, including viewing, adding, modifying, and adjusting the service provider list.

[0059] Furthermore, in the algorithm demonstration unit of the AI ​​application service module, the algorithms include video quality monitoring, speech synthesis, car search by image, license plate recognition, vehicle model-axle recognition, OCR recognition, and sensitive word recognition.

[0060] In a second aspect, the present invention further provides a method for opening an artificial intelligence application for a smart highway, the method comprising:

[0061] Construct a highway data set, collect actual highway business scenario data and upload it to the platform to form a data set asset for AI algorithm model training and development;

[0062] Build a highway model library, including pre-trained models and various algorithm models, to form model assets for AI model training acceleration or inference service deployment;

[0063] Build a highway image library, including packaged images of various algorithm operating environments, to form image assets for deployment of training environment inference services;

[0064] Build a highway algorithm library, deploy various algorithm services to the platform to form algorithm assets, and provide algorithm service interfaces;

[0065] A front-end display, connected to the highway data set, the highway model library, the highway mirror library and the highway algorithm library, for displaying a graphical interface to the user;

[0066] Deploy AI application services to implement the platform application service management function; including cloud management and edge management;

[0067] The cloud management is used to implement cloud service management, including online services and algorithm demonstrations; the online services are used for users to manage online and view the returned results of online reasoning; the algorithm demonstration is used to support multiple algorithm demonstrations, upload images, videos, audio or text according to the instructions, and view the algorithm results in real time;

[0068] The edge management enables users to manage edge application services of road sections, including section management, edge model management, edge service management and edge annotation service management; the section management is used to manage the road sections supported by the edge on the platform; the edge model management is used to implement model distribution, as well as management and deployment of edge model versions on the cloud; the edge service management is used to implement application distribution, as well as management and deployment of edge application services on the cloud; the edge annotation service management is used to implement data feedback, as well as edge annotation service details and annotation data viewing.

[0069] In a third aspect, the present invention further discloses a computer-readable storage medium storing a computer program, which implements the smart highway artificial intelligence application open method when executed by a processor.

[0070] In a fourth aspect, the present invention further discloses an electronic device, comprising a memory, a processor and a computer program, wherein the computer program is stored in the memory and can be run on the processor, and when the processor executes the computer program, the smart highway artificial intelligence application open method is implemented.

[0071] Beneficial effects:

[0072] The smart highway artificial intelligence application open platform provided by the present invention establishes a highway data set for AI algorithm model training and development; establishes a highway model library to accelerate AI model training or deploy inference services; establishes a highway mirror library for training environment inference service deployment; and implements the platform application service management function through the AI ​​application service module, supports communication with the edge after edge deployment to realize data recovery and model algorithm distribution, realizes cloud-edge collaboration, is closer to the actual scenarios of the highway industry, can continuously iterate model capabilities, and accumulate data assets.

[0073] In addition, the present invention also provides data management functions such as data annotation and data set management, computing power allocation, computing power metering and computing power management functions, algorithm online service functions, and AI asset trading functions such as algorithms and data sets. The open platform of the present invention is more functional, facing the entire life cycle of AI development and application, facilitating users to realize AI algorithm development and application in one stop, improving R&D efficiency and reducing R&D costs. BRIEF DESCRIPTION OF THE DRAWINGS

[0074] In order to make the contents of the present invention more clearly understood, the present invention is further described in detail below based on specific embodiments of the present invention in conjunction with the accompanying drawings.

[0075] Figure 1 This is a schematic diagram of the system architecture of the smart highway artificial intelligence application open platform according to Embodiment 1 of the present invention;

[0076] Figure 2 This is a schematic diagram of the structure of the AI ​​data management module in a specific embodiment of the present invention;

[0077] Figure 3 This is a schematic diagram of the structure of the AI ​​computing power management module in a specific embodiment of the present invention;

[0078] Figure 4 This is a schematic diagram of the structure of the AI ​​algorithm training control module in a specific embodiment of the present invention;

[0079] Figure 5 This is a schematic diagram of the structure of the AI ​​application service module in a specific embodiment of the present invention;

[0080] Figure 6 This is a schematic diagram of the composition of the AI ​​mall module in a specific embodiment of the present invention;

[0081] Figure 7 This is a flow chart of the open method for applying artificial intelligence to smart highways according to Example 2 of the present invention. DETAILED DESCRIPTION

[0082] The present invention will be described in detail below with reference to the accompanying drawings and in combination with embodiments. The principles and features of the present invention are described below in combination with the accompanying drawings. It should be noted that the embodiments in this application and the features in the embodiments can be combined with each other without conflict. The embodiments are only used to explain the present invention and are not used to limit the scope of the present invention.

[0083] Example 1

[0084] like Figure 1 , Figure 5 As shown, this embodiment is an open platform for intelligent highway artificial intelligence applications. The platform includes:

[0085] Highway dataset 110 collects actual highway business scenario data and uploads it to the platform to form dataset assets for AI algorithm model training and development, such as road-tunnel traffic event datasets, toll station license plate-vehicle model datasets, etc. The data types include images, videos, audio, text, etc. The datasets are collected by edge data recovery or manual upload by users.

[0086] Highway model library 120, including pre-trained models and various algorithm models to form model assets, which are used for AI model training acceleration or inference service deployment, such as traffic event detection model, video quality detection model, green pass cargo identification model, etc. The model collection method is to upload files within the platform, including weight file name, version number, model information, etc.;

[0087] The highway image library 130 includes packaged images of various algorithm operating environments to form image assets for training environment reasoning service deployment, such as traffic event detection algorithm images, video quality detection algorithm images, vehicle type recognition algorithm images, etc. The image collection method is to package the operating environment within the platform or upload the file, including the image file name, version number, image information, etc.;

[0088] Highway algorithm library 140, completes the deployment of various algorithms as services to the platform to form algorithm assets, and is used to provide algorithm service interfaces, such as traffic event detection interface, video quality detection interface, vehicle model recognition interface, license plate recognition interface, etc.;

[0089] A front-end display module 210 is connected to the highway data set 110, the highway model library 120, the highway image library 130, and the highway algorithm library 140, and is used to display a graphical interface to the user, wherein the graphical interface displays pages including AI data management, AI computing power management, AI algorithm training control, AI application services, and AI mall 350;

[0090] The AI ​​application service module 340 is connected to the AI ​​application service page of the front-end display module 210 and is used to implement the platform application service management function, including a cloud management submodule 341 and an edge management submodule 342;

[0091] The cloud management submodule 341 includes an online service unit and an algorithm demonstration unit to implement cloud service management;

[0092] The online service unit includes online service list viewing, creation, start, stop, edit, rollback, deletion, and online reasoning. Users can manage online services and view the return results of online reasoning;

[0093] The algorithm demonstration unit includes a demonstration system page that supports multiple algorithm demonstrations. Users can upload images, videos, audio or text according to the instructions and view the algorithm results in real time. It supports highway industry algorithms such as video quality monitoring, speech synthesis, image search, license plate recognition, vehicle model-axle recognition, OCR recognition, and sensitive word recognition.

[0094] The edge management submodule 342 includes a road segment management unit, an edge model management unit, an edge service management unit, and an edge annotation service management unit, to enable users to manage the application services of road segments and edges;

[0095] The road segment management unit includes road segment query, list viewing, adding, modifying, and deleting, so as to manage the road segments supported by the edge on the platform;

[0096] The edge model management unit supports the viewing and deployment of edge model versions, and manages and deploys edge model versions on the cloud platform, realizing one of the key steps of cloud-edge collaboration - model distribution;

[0097] The edge service management unit supports adding, starting, stopping, list viewing, editing, and deploying edge services, enabling management and deployment of edge application services on the cloud platform, and realizing one of the key steps of cloud-edge collaboration - application delivery;

[0098] The edge annotation service management unit supports edge annotation service details and annotation data viewing, realizing one of the key steps in cloud-edge collaboration - data feedback.

[0099] In a preferred embodiment, Figure 2 As shown, the platform also includes:

[0100] The AI ​​data management module 310 is connected to the AI ​​data management page of the front-end display module 210 and is used to implement the platform data management function, including a data set management submodule 311, a labeling task management submodule 312, and a labeling service management submodule 313;

[0101] The data set management submodule 311 includes a data set creation unit, a data set information modification unit, a data set deletion unit, a data set query unit, and a data set publishing unit. Users can operate batch data sets and get quick responses.

[0102] The dataset creation unit includes uploading images, text, audio and other data to form a dataset or importing an existing dataset. When creating a dataset, you need to enter the dataset name, data type, annotation type, label type, label group and dataset description;

[0103] The data set information modification unit includes modifying the name, data type, annotation type, label type, label group and data set description of the created data set;

[0104] The dataset deletion unit is used to delete the created datasets and supports batch deletion;

[0105] The data set query unit is used to query the data set according to the input number or name;

[0106] The dataset publishing unit is used to publish the version of the selected dataset, including the version number and publishing information;

[0107] The labeling task management submodule 312 includes a labeling task creation unit, a labeling task dispatching unit, a data labeling unit, a labeling task result review unit, a labeling task result viewing unit, and a labeling task batch deletion unit, and users can manage labeling tasks;

[0108] The labeling task creation unit creates a labeling task by inputting a task name and selecting a data set;

[0109] The annotation task dispatching unit supports splitting the created annotation tasks and dispatching them to multiple users;

[0110] The data annotation unit supports users to annotate the data set of the annotation task. The annotation methods include manual annotation and automatic annotation. Manual annotation supports users to annotate data through a graphical interface. Automatic annotation automatically annotates the selected data set by selecting an annotation service. The annotation content includes image target area and type, image segmentation, video target tracking, speech recognition, text classification and word segmentation, etc.

[0111] The labeling task result review unit allows users with review authority to check completed labeling tasks and return unqualified data for re-labeling;

[0112] The labeling task result viewing unit allows users to view data labeling results;

[0113] The annotation task deletion unit supports batch deletion of annotation tasks;

[0114] The annotation service management submodule 313 includes an annotation service creation unit, an annotation service start and stop unit, an annotation service editing unit, and an annotation service batch deletion unit. Users can manage annotation tasks to achieve automatic annotation.

[0115] The annotation service creation unit supports creating annotation services in a graphical interface by entering the service name and selecting the model type, algorithm type, model version, image version, node type and specification, and service quantity;

[0116] The labeling service start / stop unit includes starting / stopping operations on the selected labeling service;

[0117] The annotation service editing unit includes modification of the selected annotation service, and supports modification of model type, algorithm type, model version, image version, node type and specification, and service quantity;

[0118] The batch deletion unit of the annotation service supports the deletion of batch annotation services;

[0119] In a preferred embodiment, Figure 3 As shown, the platform also includes:

[0120] The AI ​​computing power management module 320 is connected to the AI ​​computing power management page of the front-end display module 210 and is used to implement the platform computing power management function, including a cluster status query unit 321, a resource specification management unit 322, a node label management unit 323, and a resource management unit 324;

[0121] The cluster status query unit 321 supports viewing the platform k8s cluster status and realizes monitoring of the health status of the platform cluster;

[0122] The resource specification management unit 322 includes creating, editing, and deleting platform container resources to achieve platform container resource management;

[0123] The node label management unit 323 includes modifying, deactivating, and deleting platform node labels to achieve management of platform computing nodes;

[0124] The resource management unit 324 includes a platform GPU virtualization subunit, a permission and quota management subunit, a resource pool creation / deletion subunit, and a computing power measurement subunit to achieve platform computing resource management;

[0125] The platform GPU virtualization subunit supports splitting a GPU into multiple virtual GPUs according to quantity and providing them to different users;

[0126] The authority and quota management subunit supports user authority management to achieve user-level management of computing power, and supports user quota management to achieve on-demand allocation of computing power;

[0127] The resource pool creation / deletion subunit supports administrators to create or delete platform resource pools;

[0128] The computing power metering subunit supports statistics on the total amount of computing power actually used by users to support billing operations.

[0129] In a preferred embodiment, Figure 4 As shown, the platform also includes:

[0130] The AI ​​algorithm training control module 330 is connected to the AI ​​algorithm training control page of the front-end display module 210, and is used to implement platform algorithm training and management functions, including an image management submodule 331, a model management submodule 332, an algorithm development submodule 333, and a terminal management submodule 334;

[0131] The image management submodule 331 includes an image uploading unit, an image editing unit, and an image batch deleting unit to implement the management of platform images;

[0132] The image upload unit supports algorithm development users to upload algorithm running image files to the platform;

[0133] The image editing unit supports algorithm developers to modify image information, including image name, image purpose, image address, and image version number;

[0134] The image batch deletion unit supports users to delete images in batches;

[0135] The model management submodule 332 includes a model weight management unit and a model optimization task management unit to manage the platform model weight files and optimization tasks;

[0136] The model weight management unit includes model weight list viewing, uploading, and version management, which enables model weight file information viewing, supports users to upload model weights to the platform and perform version management;

[0137] The model optimization task management unit supports inputting task names and descriptions in the graphical interface, selecting optimization types, optimization algorithms, models, and data sets to create model optimization tasks, and supports starting, stopping, modifying, and deleting model optimization tasks.

[0138] The algorithm development unit 333 includes a notebook management subunit, an algorithm management subunit, and a training task management subunit, and supports algorithm development users to implement customized algorithm development;

[0139] The Notebook management subunit includes Notebook creation, start, stop, delete, and view, which supports users to manage Notebook algorithm running scripts;

[0140] The algorithm management subunit includes algorithm upload, modification, editing, downloading, and deletion, supporting users to implement algorithm code management;

[0141] The training task management subunit includes training task creation, viewing, stopping, and deletion. Users can independently create and start and stop training tasks, and monitor resources and view logs for training tasks.

[0142] The terminal management unit 334 includes terminal overview and remote connection functions to implement user platform remote connection management. Users can create, save and stop, restart, delete remote connections, and view remote connection related information.

[0143] In a preferred embodiment, Figure 6 As shown, the platform also includes:

[0144] The AI ​​mall module 350 is connected to the AI ​​mall page of the front-end display module 210 and is used to realize the online transaction function of platform data and algorithm assets, including a mall page submodule 351, a commodity management submodule 352 and a service provider management submodule 353;

[0145] The mall page submodule 351 includes the mall homepage, login page, and product details page, which implements data algorithm product display and details viewing;

[0146] The product management submodule 352 includes product addition, editing, listing application, listing review, and delisting, realizing data algorithm product management. The administrator can manage the entire process of the mall products;

[0147] The service provider management submodule 353 includes service provider list viewing, adding, modifying, and member adjustment, and implements third-party service provider member management within the platform.

[0148] Example 2

[0149] like Figure 7 As shown, this embodiment also provides a method for opening an artificial intelligence application for a smart highway, the method comprising:

[0150] Construct highway dataset 410, collect actual highway business scenario data and upload it to the platform to form dataset assets for AI algorithm model training and development;

[0151] Build a highway model library 420, including pre-trained models and various algorithm models, to form model assets for AI model training acceleration or inference service deployment;

[0152] Build a highway image library 430, including various algorithm runtime environment packaged images, forming image assets for training environment reasoning service deployment;

[0153] Build a highway algorithm library 440, deploy various algorithms as services to the platform to form algorithm assets, and provide algorithm service interfaces;

[0154] Deploy AI application service 450 to implement platform application service management functions, including cloud management and edge management;

[0155] Cloud management is used to implement cloud service management, including online services and algorithm demonstrations; online services are used for user online management and online reasoning to view the returned results; algorithm demonstration is used to support multiple algorithm demonstrations, upload images, videos, audio or text according to the instructions, and view the algorithm results in real time;

[0156] Edge management enables users to manage edge application services of road sections, including road section management, edge model management, edge service management, and edge annotation service management; road section management is used to manage road sections supported by the edge on the platform; edge model management is used to distribute models, manage and deploy edge model versions on the cloud; edge service management is used to distribute applications, manage and deploy edge application services on the cloud platform; edge annotation service management is used to transmit data, and view edge annotation service details and annotation data.

[0157] The front-end display 460 is connected to the highway data set 110, the highway model library 120, the highway image library 130 and the highway algorithm library 140, and is used to display a graphical interface to the user;

[0158] The present invention also provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the above-mentioned smart highway artificial intelligence application open method.

[0159] The present invention also provides an electronic device, including a memory, a processor and a computer program. The computer program is stored in the memory and can be run on the processor. When the processor executes the computer program, the above-mentioned smart highway artificial intelligence application open method is implemented.

[0160] Obviously, the above embodiments are merely examples for the purpose of clear explanation, and are not intended to limit the implementation methods. For those skilled in the art, other different forms of changes or modifications can be made based on the above description. It is not necessary and impossible to list all the implementation methods here. The obvious changes or modifications derived therefrom are still within the scope of protection of the invention.

Claims

1. A smart highway artificial intelligence application open platform, characterized by: The platform includes: Highway data set, which collects actual highway business scenario data and uploads it to the platform to form data set assets for AI algorithm model training and development; Highway model library, including pre-trained models and various algorithm models, forming model assets for AI model training acceleration or inference service deployment; The highway image library includes packaged images of various algorithm operating environments to form image assets for deployment of training environment inference services; Highway algorithm library, which deploys various algorithms as services to the platform to form algorithm assets and is used to provide algorithm service interfaces; A front-end display module, connected to the highway data set, the highway model library, the highway mirror library and the highway algorithm library, for displaying a graphical interface to the user; AI application service module, used to implement the platform application service management function, including cloud management submodule and edge management submodule; The cloud management submodule is used to implement cloud service management, including an online service unit and an algorithm demonstration unit; the online service unit is used for user online management and online reasoning to view the returned results; the algorithm demonstration unit is used to support multiple algorithm demonstrations, upload images, videos, audio or text according to the instructions, and view the algorithm results in real time; The edge management submodule is used to implement user management of road segment edge application services, including a road segment management unit, an edge model management unit, an edge service management unit and an edge annotation service management unit; the road segment management unit is used to implement the platform's management of road segments supported by the edge; the edge model management unit is used to implement model distribution, as well as management and deployment of edge model versions on the cloud; the edge service management unit is used to implement application distribution, as well as management and deployment of edge application services on the cloud; the edge annotation service management unit is used to implement data feedback, as well as edge annotation service details and annotation data viewing.

2. The platform according to claim 1, characterized in that The platform also includes an AI data management module, which includes: The data set management submodule is used for users to operate batch data sets and get quick responses, including a data set creation unit, a data set information modification unit, a data set deletion unit, a data set query unit, and a data set publishing unit; The annotation task management submodule is used for users to manage annotation tasks. The annotation methods include manual annotation and automatic annotation. The manual annotation is performed by users through a graphical interface to annotate data. The automatic annotation is performed by selecting an annotation service to automatically annotate the selected data set. The annotation task management submodule includes an annotation task creation unit, an annotation task dispatching unit, a data annotation unit, an annotation task result review unit, an annotation task result viewing unit, and an annotation task batch deletion unit. The annotation service management submodule is used for users to manage annotation tasks to achieve the automatic annotation; the annotation service management submodule includes an annotation service creation unit, an annotation service start and stop unit, an annotation service editing unit and an annotation service batch deletion unit.

3. The platform according to claim 2, characterized in that: In the data set management submodule, The data set creation unit includes uploading images, texts, and audio data to form a data set or importing an existing data set, and inputting a data set name, data type, annotation type, label type, label group, and data set description during creation; The data set information modification unit includes modifying the name, data type, annotation type, label type, label group and data set description of the created data set; The data set deletion unit is used to delete the created data set; The data set query unit queries the data set according to the input number or name; The data set publishing unit publishes a version of the selected data set, including a version number and publishing information; In the annotation task management submodule, The labeling task creation unit creates a labeling task by inputting a task name and selecting a data set; The labeling task dispatching unit splits the created labeling task and dispatches it to multiple users; The data annotation unit is used for users to annotate the data set to which the annotation task belongs, and the annotation content includes image target area and type, image segmentation, video target tracking, speech recognition, and text classification and segmentation; The labeling task result review unit is used for users with review authority to review completed labeling tasks and re-label unqualified data; The labeling task result viewing unit is used for users to view data labeling results; The labeling task batch deletion unit is used to delete the labeling tasks in batches; In the annotation service management submodule, The annotation service creation unit is used to create an annotation service by inputting a service name and selecting a model type, an algorithm type, a model version, an image version, a node type and specification, and a service quantity in a graphical interface; The annotation service start and stop unit is used to start and stop the selected annotation service; The annotation service editing unit is used to modify the selected annotation service, including modifying the model type, algorithm type, model version, image version, node type and specification, and service quantity; The annotation service batch deletion unit is used to delete the batch annotation services.

4. The platform according to claim 1, characterized in that: The platform also includes an AI computing power management module, which includes: The cluster status query unit is used to check the status of the platform cluster and monitor the health status of the platform cluster; Resource specification management unit, used for managing platform container resources; Node label management unit, used to manage platform computing nodes; A resource management unit is used to manage platform computing resources, and the resource management unit includes: The platform GPU virtualization subunit is used to divide a GPU into multiple virtual GPUs according to the quantity and provide them to different users; The authority and quota management subunit is used for user authority management to achieve user-level management of computing power, and user quota management to achieve on-demand allocation of computing power; The resource pool creation / deletion subunit is used by administrators to create or delete platform resource pools; The computing power metering subunit is used to count the total amount of computing power actually used by users for billing operations.

5. The platform according to claim 1, characterized in that: The platform also includes an AI algorithm training control module, which includes: The image management submodule is used to manage the platform images, including image uploading, image editing and image batch deletion; The model management submodule is used to manage the platform model weight files and optimization tasks, including: Model weight management unit, including model weight list viewing, uploading, and version management, is used to view model weight file information, and users upload model weights to the platform and perform version management; A model optimization task management unit, used to input a task name and a task description in the graphical interface, select an optimization type, an optimization algorithm, a model and a data set to create a model optimization task, and start, stop, modify and delete the model optimization task; The algorithm development unit is used by algorithm development users to implement custom algorithm development. The algorithm development unit includes: Notebook management subunit, used by users to manage Notebook algorithm running scripts; Algorithm management subunit, used by users to implement algorithm code management; The training task management subunit is used by users to independently create and start and stop training tasks, and to monitor resources and view logs of training tasks; The terminal management unit is used for users to manage remote connections on the platform.

6. The platform according to claim 1, characterized in that: The platform also includes an AI mall module, which includes a mall page submodule, a product management submodule, and a service provider management submodule; The mall page submodule is used for data algorithm product display and detail viewing, including the mall homepage, login page, and product detail page; The product management submodule is used by administrators to manage the entire process of the mall products, including product addition, editing, listing application, listing review, and delisting; The service provider management submodule is used to manage third-party service provider members on the platform, including viewing, adding, modifying, and adjusting the service provider list.

7. The platform according to claim 1, characterized in that In the algorithm demonstration unit of the AI ​​application service module, the algorithms include video quality monitoring, speech synthesis, car search by image, license plate recognition, vehicle model-axle recognition, OCR recognition, and sensitive word recognition.

8. A method for opening the application of artificial intelligence in a smart highway, characterized in that: The method comprises: Construct a highway data set, collect actual highway business scenario data and upload it to the platform to form a data set asset for AI algorithm model training and development; Build a highway model library, including pre-trained models and various algorithm models, to form model assets for AI model training acceleration or inference service deployment; Build a highway image library, including packaged images of various algorithm operating environments, to form image assets for deployment of training environment inference services; Build a highway algorithm library, deploy various algorithm services to the platform to form algorithm assets, and provide algorithm service interfaces; A front-end display, connected to the highway data set, the highway model library, the highway mirror library and the highway algorithm library, for displaying a graphical interface to the user; Deploy AI application services to implement the platform application service management function; including cloud management and edge management; The cloud management is used to implement cloud service management, including online services and algorithm demonstrations; the online services are used for users to manage online and view the returned results of online reasoning; the algorithm demonstration is used to support multiple algorithm demonstrations, upload images, videos, audio or text according to the instructions, and view the algorithm results in real time; The edge management enables users to manage edge application services of road sections, including section management, edge model management, edge service management and edge annotation service management; the section management is used to manage the road sections supported by the edge on the platform; the edge model management is used to implement model distribution, as well as management and deployment of edge model versions on the cloud; the edge service management is used to implement application distribution, as well as management and deployment of edge application services on the cloud; the edge annotation service management is used to implement data feedback, as well as edge annotation service details and annotation data viewing.

9. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the smart highway artificial intelligence application opening method described in claim 8 is implemented.

10. An electronic device comprising a memory, a processor and a computer program, wherein the computer program is stored in the memory and can be run on the processor, wherein: When the processor executes the computer program, the smart highway artificial intelligence application opening method as described in claim 8 is implemented.

Images