Image detection method and related equipment based on network state adaptive perception

Abstract

The embodiment of the present application provides an image detection method and related equipment based on adaptive perception of network status, which belongs to the field of image detection technology. The method includes: predicting the network status between the terminal and the cloud server within a preset time period. If the network status is characterized as a low-quality state, the target image is detected by a first detection model deployed on the terminal. If the network status is characterized as a high-quality state, the target image is sent to the cloud server, and the target image is detected by a second detection model deployed on the cloud server to obtain a second detection result; and then the second detection result is sent to the terminal. Since the image detection process is continuous, some target images are detected by the second detection model when the network status is good, which improves the accuracy of image detection. Some target images are detected by the first detection model when the network status is poor, which ensures the accuracy of image detection while improving the detection success rate.

Description

Technical Field

[0001] The present application relates to the field of image detection technology, and in particular to an image detection method based on adaptive perception of network status and related equipment. Background Art

[0002] In the field of computer vision, image detection technology is used to detect target objects in images, determine the semantic categories of these target objects, and calibrate the positions of target objects in images. In related technologies, a common image detection method is to perform detection through a high-precision detection model deployed on a cloud server. This type of method generally trains a complex detection model with high performance and deploys it on a cloud server. In the application stage, after the terminal captures the image, it needs to transmit it to the cloud server through the network. Then, after the cloud server completes the detection, it transmits the results back to the terminal. The advantage of this type of method is high accuracy, and the disadvantage is that it relies on the network and is prone to detection failure. Therefore, how to improve the detection success rate while ensuring the accuracy of image detection has become a technical problem that needs to be solved urgently. Summary of the invention

[0003] The main purpose of the embodiments of the present application is to propose an image detection method and related equipment based on adaptive perception of network status, aiming to ensure the accuracy of image detection while improving the detection success rate.

[0004] To achieve the above-mentioned purpose, a first aspect of an embodiment of the present application proposes an image detection method based on adaptive perception of network status, which is applied to a terminal. The method includes:

[0005] Get the target image;

[0006] Predicting the network status between the terminal and the cloud server within a preset time period; wherein the network status includes a low-quality status and a high-quality status;

[0007] If the network state is characterized as a low-quality state, object detection is performed on the target image through a preset first detection model to obtain a first detection result; wherein the first detection model is a detection model deployed on the terminal;

[0008] If the network state is characterized as a high-quality state, the target image is sent to the cloud server, and the cloud server performs object detection on the target image through a preset second detection model to obtain a second detection result; wherein the number of processing layers included in the second detection model is greater than the number of processing layers included in the first detection model;

[0009] Receive the second detection result returned by the cloud server.

[0010] In some embodiments, the network state further includes a medium quality state, and before acquiring the target image, the method further includes:

[0011] Acquire an end-cloud integrated detection model, wherein the end-cloud integrated detection model includes multiple processing layers;

[0012] determining a target partitioning layer among the plurality of processing layers;

[0013] Among the multiple processing layers, a terminal-side operation sub-model is formed according to the processing layers between the first processing layer and the target division layer, and the terminal-side operation sub-model is deployed on the terminal;

[0014] Among the plurality of processing layers, a cloud operation sub-model is formed according to the processing layers from the processing layer after the target division layer to the last processing layer, and the cloud operation sub-model is deployed on the cloud server;

[0015] After obtaining the network status between the terminal and the cloud server, the method further includes:

[0016] If the network state is characterized as a medium quality state, the target image is input into the end-side running sub-model for object detection to obtain intermediate layer features;

[0017] The intermediate layer features are sent to the cloud server, and the cloud server performs object detection on the intermediate layer features through the cloud operation sub-model to obtain a third detection result;

[0018] Receive the third detection result returned by the cloud server.

[0019] In some embodiments, the medium quality state includes a first quality state and a second quality state, and a network state corresponding to the first quality state is worse than a network state corresponding to the second quality state;

[0020] The determining of a target partitioning layer among the plurality of processing layers comprises:

[0021] If the medium quality state is characterized as the first quality state, then, among the multiple processing layers, determining the Nth processing layer as the target partitioning layer;

[0022] If the medium quality state is characterized as the second quality state, then among the multiple processing layers, the Mth processing layer is determined as the target division layer; wherein N and M are both positive integers, and N is greater than M.

[0023] In some embodiments, the sending of the intermediate layer features to the cloud server, and the cloud server performing object detection on the intermediate layer features through the cloud running sub-model to obtain a third detection result, includes:

[0024] The intermediate layer features are compressed by a preset coding and compression module to obtain intermediate layer compressed features; wherein the intelligent compression model includes a coding and compression module and a decoding module, the coding and compression module is deployed on the terminal, and the decoding module is deployed on the cloud server;

[0025] The intermediate layer compression features are sent to the cloud server, and the cloud server decodes the intermediate layer compression features through the decoding module to obtain the intermediate layer decoded features, and performs object detection on the intermediate layer decoded features through the cloud running sub-model to obtain the third detection result.

[0026] In some embodiments, before compressing the intermediate layer features by a preset encoding compression module to obtain the intermediate layer compressed features, the method further includes:

[0027] The end-cloud integrated detection model is trained, including:

[0028] Get image samples;

[0029] Inputting the image sample into the end-side running sub-model to obtain intermediate features of the sample;

[0030] Inputting the sample intermediate features into the encoding compression module to obtain sample intermediate compressed features;

[0031] Inputting the sample intermediate compression feature into the decoding module to obtain the sample intermediate decoding feature;

[0032] Inputting the sample intermediate decoding features into the cloud-based running sub-model to obtain a sample detection result;

[0033] Calculating loss data according to the sample detection results and preset detection labels to obtain target loss data;

[0034] Parameters of the end-cloud integrated detection model are adjusted according to the target loss data to obtain the trained end-cloud integrated detection model.

[0035] In some embodiments, before acquiring the target image, the method further includes:

[0036] Constructing the first detection model includes:

[0037] Acquire an intelligent pruning model; wherein the intelligent pruning model includes the initial detection model and a parameter importance evaluation model, the parameter importance evaluation model is used to evaluate the importance of parameters of each processing layer in the initial detection model, and the initial detection model includes multiple processing layers;

[0038] Obtain training image data;

[0039] Inputting the training image data into the initial detection model to perform object detection, and obtaining output features of each processing layer;

[0040] Inputting the parameters of each processing layer in the initial detection model and each output feature into the parameter importance evaluation model for importance evaluation to obtain the importance score of each parameter;

[0041] Determine the parameter whose importance score is lower than a preset threshold as the target parameter, and set the target parameter in the initial detection model to zero;

[0042] The parameters of the initial detection model are adjusted based on a preset loss function to obtain the first detection model.

[0043] In some embodiments, predicting the network status between the terminal and the cloud server within a preset time period includes:

[0044] Acquiring interaction data between the terminal and the cloud server;

[0045] Parsing the interaction data to obtain a plurality of first data indicators;

[0046] Inputting the plurality of the first data indicators into a preset network status prediction model to perform network status prediction, and obtaining a plurality of second data indicators within a preset time period;

[0047] The network status is obtained according to a plurality of the second data indicators and a network status mapping table, wherein the network status mapping table contains network statuses corresponding to different combinations of second data indicators.

[0048] To achieve the above-mentioned purpose, a second aspect of an embodiment of the present application proposes an image detection device based on adaptive perception of network status, the device comprising:

[0049] An acquisition module, used for acquiring a target image;

[0050] A prediction module, used to predict the network status between the terminal and the cloud server within a preset time period; wherein the network status includes a low-quality status and a high-quality status;

[0051] A detection module, configured to perform object detection on the target image through a preset first detection model to obtain a first detection result if the network state is characterized as a low-quality state; wherein the first detection model is a detection model deployed on the terminal;

[0052] a sending module, configured to send the target image to the cloud server if the network state is characterized as a high-quality state, and the cloud server performs object detection on the target image through a preset second detection model to obtain a second detection result; wherein the number of processing layers included in the second detection model is greater than the number of processing layers included in the first detection model;

[0053] A receiving module is used to receive the second detection result returned by the cloud server.

[0054] To achieve the above objectives, a third aspect of an embodiment of the present application proposes an electronic device, which includes a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the method described in the first aspect is implemented.

[0055] To achieve the above-mentioned purpose, the fourth aspect of an embodiment of the present application proposes a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the method described in the first aspect is implemented.

[0056] The image detection method and related equipment based on adaptive perception of network status proposed in the present application predict the network status between the terminal and the cloud server within a preset time period. If the network status is characterized as a low-quality state, the target image is detected by the first detection model deployed on the terminal. If the network status is characterized as a high-quality state, the target image is sent to the cloud server, and the second detection model deployed on the cloud server performs high-precision object detection on the target image to obtain a second detection result; the cloud server then sends the second detection result to the terminal. Since the image detection process is continuous, some target images are detected by the second detection model when the network status is good, thereby improving the accuracy of image detection. When the network status of some target images is poor, the target images cannot be sent to the cloud server. In order to prevent detection failure, the first detection model is used for detection. While ensuring the accuracy of image detection, the detection success rate is improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] Figure 1 is an optional flow chart of an image detection method based on network state adaptive perception provided in an embodiment of the present application;

[0058] Figure 2 is a flow chart of an image detection method based on adaptive perception of network status provided by a second embodiment of the present application;

[0059] Figure 3 is a flowchart of an image detection method based on adaptive perception of network status provided by a third embodiment of the present application;

[0060] Figure 4 yes Figure 2 Flow chart of step S202 in FIG.

[0061] Figure 5 yes Figure 3 Flow chart of step S302 in FIG.

[0062] Figure 6 is a flowchart of an image detection method based on adaptive perception of network status provided by a fourth embodiment of the present application;

[0063] Figure 7 is a flowchart of an image detection method based on adaptive perception of network status provided by a fifth embodiment of the present application;

[0064] Figure 8 yes Figure 1 Flow chart of step S102 in FIG.

[0065] Fig. 9 is a structural schematic diagram of an image detection device based on adaptive perception of network status provided in an embodiment of the present application;

[0066] Fig.10 It is a schematic diagram of the hardware structure of the electronic device provided in the embodiment of the present application. DETAILED DESCRIPTION

[0067] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0068] It should be noted that, although the functional modules are divided in the device schematic diagram and the logical order is shown in the flowchart, in some cases, the steps shown or described may be performed in a different order than the module division in the device or the order in the flowchart. The terms "first", "second", etc. in the specification, claims and the above drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence.

[0069] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by those skilled in the art to which this application belongs. The terms used herein are only for the purpose of describing the embodiments of this application and are not intended to limit this application.

[0070] First, some nouns involved in this application are analyzed:

[0071] First detection model: The first detection model is deployed on the terminal. Due to the limitation of the hardware computing resources of the terminal, only a lightweight detection model can be deployed. Therefore, the first detection model is a lightweight detection model.

[0072] Second detection model: The second detection model is deployed on a cloud server. The hardware computing resources of the cloud server can support complex detection models, so the second detection model is a high-precision detection model.

[0073] A common image detection method is to perform detection through a high-precision detection model deployed on a cloud server. This type of method generally trains a complex detection model with high performance and deploys it on a cloud server. In the application stage, after the terminal collects the image, it needs to transmit it to the cloud server through the network. Then, after the cloud server completes the detection, it transmits the result back to the terminal. The advantage of this type of method is high accuracy, but the disadvantage is that it relies on the network and is prone to detection failure. Therefore, it is impossible to improve the detection success rate while ensuring the accuracy of image detection.

[0074] Based on this, the embodiment of the present application provides an image detection method and related equipment based on adaptive perception of network status, which aims to obtain the network status within a preset time period in the future. When the network status is characterized as a low-quality state, the target image is detected by a first detection model deployed on the terminal. If the network status is characterized as a high-quality state, the target image is sent to the cloud server, and the target image is detected by a second detection model deployed on the cloud server, and the detection result is returned to the terminal. That is, in the case of a weak network or no network, when the target image cannot be sent to the cloud server for high-precision detection, the target image is first detected by the first detection model deployed on the terminal. After the network status is restored to a high-quality state, the target image is detected by the second detection model deployed on the cloud server, which ensures the accuracy of image detection while improving the detection success rate.

[0075] The image detection method based on adaptive perception of network status and related equipment provided in the embodiments of the present application are specifically illustrated through the following embodiments. First, the image detection method based on adaptive perception of network status in the embodiments of the present application is described.

[0076] The embodiments of the present application can acquire and process relevant data based on artificial intelligence technology. Artificial Intelligence (AI) is the theory, method, technology and application system that uses digital computers or machines controlled by digital computers to simulate, extend and expand human intelligence, perceive the environment, acquire knowledge and use knowledge to obtain the best results.

[0077] AI basic technologies generally include sensors, dedicated AI chips, cloud computing, distributed storage, big data processing technology, operation / interaction systems, mechatronics, etc. AI software technologies mainly include computer vision technology, robotics technology, biometrics technology, speech processing technology, natural language processing technology, and machine learning / deep learning.

[0078] The image detection method based on adaptive perception of network status provided in the embodiment of the present application is applied to the terminal, and can also be software running in the terminal. In some embodiments, the terminal can be a smart phone, a tablet computer, a laptop computer, a desktop computer, etc.; the server side can be configured as an independent physical server, or a server cluster or distributed system composed of multiple physical servers, and can also be configured as a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, CDN, and big data and artificial intelligence platforms; the software can be an application that implements the image detection method based on adaptive perception of network status, etc., but is not limited to the above forms.

[0079] The present application can be used in many general or special computer system environments or configurations. For example: personal computers, server computers, handheld or portable devices, tablet devices, multiprocessor systems, microprocessor-based systems, set-top boxes, programmable consumer electronics, network PCs, minicomputers, mainframe computers, distributed computing environments including any of the above systems or devices, etc. The present application can be described in the general context of computer-executable instructions executed by a computer, such as program modules. Generally, program modules include routines, programs, objects, components, data structures, etc. that perform specific tasks or implement specific abstract data types. The present application can also be practiced in distributed computing environments, in which tasks are performed by remote processing devices connected through a communication network. In a distributed computing environment, program modules can be located in local and remote computer storage media including storage devices.

[0080] Please refer to Figure 1 , Figure 1 It is an optional flow chart of the image detection method based on adaptive perception of network status provided in an embodiment of the present application. The image detection method based on adaptive perception of network status is applied to a terminal. Figure 1 The method may include but is not limited to steps S101 to S105.

[0081] Step S101, acquiring a target image;

[0082] Step S102, predicting the network status between the terminal and the cloud server within a preset time period; wherein the network status includes a low-quality status and a high-quality status;

[0083] Step S103: if the network status is characterized as a low-quality status, object detection is performed on the target image through a preset first detection model to obtain a first detection result; wherein the first detection model is a detection model deployed on the terminal;

[0084] Step S104: if the network state is characterized as a high-quality state, the target image is sent to the cloud server, and the cloud server performs object detection on the target image through a preset second detection model to obtain a second detection result; wherein the number of processing layers included in the second detection model is greater than the number of processing layers included in the first detection model;

[0085] Step S105, receiving the second detection result returned by the cloud server.

[0086] In step S101 of some embodiments, the target image is an image obtained by photographing an object in real time through an image collector installed on the terminal. The photographed object is moving, and the terminal can be stationary or moving. The scene where the terminal is moving is an aerial refueling scene, and the terminal is a tanker, and the tanker needs to refuel the refueled aircraft. At this time, an image collector is installed on the tanker, and the object photographed is the refueled aircraft. The image collector on the tanker photographs the refueled aircraft in real time to obtain a target image. The refueled aircraft is detected and tracked according to the target image, so as to achieve aerial docking and complete aerial refueling.

[0087] The scenario where the terminal is stationary includes a vehicle detection scenario. In this case, the terminal is a plurality of camera modules installed on the roadside. Each camera module includes an image collector, and the object photographed is the vehicle that needs to be detected and tracked. When the vehicle drives out of the shooting range of the current camera module and enters the shooting range of the next camera module, the next camera module will take a real-time photo of the vehicle. Therefore, in this scenario, the terminal is stationary.

[0088] In step S102 of some embodiments, the terminal analyzes and obtains the current network status by performing random data interaction with the cloud server, and predicts the network status within a preset time period in the future based on the current network status. When the network status is in a low-quality state, it means that the current network environment is a weak network or no network environment. It can be seen from the above scenario that no matter whether the terminal is changing or stationary, the network environment in which the terminal is located changes in real time. Therefore, it is necessary to predict the network status between the terminal and the cloud server within a preset time period every preset period. Among them, the future prediction time period can be within the next 1s, and the prediction period can also be 1s, both of which can be adjusted according to the speed of network changes.

[0089] In step S103 of some embodiments, when the network status is a low-quality status, it indicates that the current network environment is a weak network or no network environment. At this time, the target image cannot be successfully sent to the cloud server. In order to successfully perform object detection, the target image is subjected to object detection by a first detection model deployed on the terminal to obtain a first detection result.

[0090] In step S104 of some embodiments, the processing layers of the detection model include convolutional layers, pooling layers, global pooling layers, and detection layers, etc. The more complex the network is, the more each processing layer is, and the higher the accuracy rate of the corresponding detection result is. Therefore, the number of processing layers included in the second detection model is greater than the number of processing layers included in the first detection model. When the network state is a high-quality state, it means that the current network environment is very good and the target image can be quickly sent to the cloud server. The second detection model deployed on the cloud server performs high-precision object detection on the target image to obtain a second detection result.

[0091] In step S105 of some embodiments, the cloud server returns the second detection result to the terminal, so that the terminal can perform subsequent processing according to the second detection result.

[0092] In actual applications, the network status between the terminal and the cloud server within a preset time period is predicted at every preset period. When the network status changes from a high-quality state to a low-quality state, the second detection model is switched to the first detection model for object detection. When the network status changes from a low-quality state to a high-quality state, the first detection model is switched to the second detection model for object detection.

[0093] In steps S101 to S105 shown in the embodiment of the present application, by predicting the network status between the terminal and the cloud server within a preset time period, if the network status is characterized as a low-quality state, the target image is subjected to object detection through the first detection model deployed on the terminal. If the network status is characterized as a high-quality state, the target image is sent to the cloud server, and the second detection model deployed on the cloud server performs object detection on the target image to obtain a second detection result; the cloud server then sends the second detection result to the terminal. Since the image detection process is continuous, some target images are detected by the second detection model when the network status is good, thereby improving the accuracy of image detection. Some target images are detected by the first detection model when the network status is poor, thereby ensuring the accuracy of image detection and improving the detection success rate.

[0094] See also Figure 2In some embodiments, the network state further includes a medium quality state. Before step S101, the image detection method based on adaptive perception of network state may also include but is not limited to steps S201 to S204:

[0095] Step S201, obtaining a terminal-cloud integrated detection model, where the terminal-cloud integrated detection model includes multiple processing layers;

[0096] Step S202, determining a target partitioning layer among multiple processing layers;

[0097] Step S203, among the multiple processing layers, forming a terminal-side operation sub-model according to the processing layers between the first processing layer and the target division layer, and deploying the terminal-side operation sub-model on the terminal;

[0098] Step S204, among the multiple processing layers, a cloud operation sub-model is formed according to the processing layers from the processing layer after the target division layer to the last processing layer, and the cloud operation sub-model is deployed on the cloud server;

[0099] In step S201 of some embodiments, the original end-cloud integrated detection model is trained by the total loss function to obtain a trained end-cloud integrated detection model. The total loss function includes regression loss, classification loss, and end-side parameter sparse loss. Before model training, it is determined which processing layers are the layers running on the terminal and which processing layers are the layers running on the cloud server. The end-side parameter sparse loss is used to make the L1 norm of the parameters of the layer running on the terminal as small as possible, to reduce the complexity of the layer running on the terminal, to reduce the computing power requirements of the terminal, and to be able to run on more simple terminals.

[0100] In step S202 of some embodiments, the target division layer can be set according to demand, and multiple end-cloud integrated detection models can be trained, each end-cloud integrated detection model has a different target division layer.

[0101] In steps S203 to S204 of some embodiments, if the end-cloud integrated detection model has a total of 10 layers, and the 5th layer is the target division layer, then the processing layers between the 1st layer and the 5th layer constitute the end-side operation sub-model. The processing layers between the 6th layer and the 10th layer constitute the cloud-side operation sub-model. After the end-cloud integrated detection model is trained, the end-side operation sub-model is deployed on the terminal, and the cloud-side operation sub-model is deployed on the cloud server.

[0102] See also Figure 3 In some embodiments, after step S102, the image detection method based on network state adaptive perception may also include but is not limited to steps S301 to S303:

[0103] Step S301: If the network status is characterized as a medium quality status, the target image is input to the client-side running sub-model for object detection to obtain intermediate layer features;

[0104] Step S302, sending the intermediate layer features to the cloud server, and the cloud server performs object detection on the intermediate layer features by running the sub-model in the cloud to obtain a third detection result;

[0105] Step S303, receiving the third detection result returned by the cloud server.

[0106] In step S301 of some embodiments, when the network status is characterized as a medium quality status, the target image is sent to the cloud server at a slow speed due to the large amount of data contained in the target image. Therefore, the terminal side may first run the sub-model to detect the target image and output the intermediate layer features.

[0107] In some embodiments, in step S302 to step S303, the amount of data contained in the intermediate layer is less than the amount of data contained in the target image. The intermediate layer features are sent to the cloud server, and the cloud server performs high-precision object detection on the intermediate layer features through the cloud running sub-model to obtain a third detection result, and returns the third detection result to the terminal.

[0108] In steps S301 to S303 shown in this embodiment, when the network status is medium quality, it is impossible to quickly send the target image with a large amount of data to the cloud server. First, the end-side runs the sub-model to detect the target object and output the intermediate layer features. The intermediate layer features are quickly sent to the cloud server, and the intermediate layer features are continued to be detected by the high-precision cloud-based sub-model, which can improve the speed of image detection and the detection accuracy.

[0109] See also Figure 4 In some embodiments, the medium quality state includes a first quality state and a second quality state, and the network state corresponding to the first quality state is worse than the network state corresponding to the second quality state. Then step S202 may include but is not limited to steps S401 to S402:

[0110] Step S401, if the medium quality state is characterized as the first quality state, then among the multiple processing layers, the Nth processing layer is determined as the target partitioning layer;

[0111] Step S402: If the medium quality state is characterized as the second quality state, then among the multiple processing layers, the Mth processing layer is determined as the target partitioning layer; wherein N and M are both positive integers, and N is greater than M.

[0112] In steps S401 to S402 of some embodiments, as the network runs, the resolution of its features decreases. Therefore, when the terminal contains more processing layers, the terminal runs more links, the less data needs to be transmitted to the cloud server, and the less dependence on the network. It is more suitable for situations where the network is not so good. On the contrary, when the network is good, more data can be transmitted to the cloud server. Therefore, the terminal operation link can contain fewer processing layers, so that the cloud operation sub-model contains more processing layers, thereby improving the detection accuracy.

[0113] Specifically, the network state corresponding to the first quality state is worse than the network state corresponding to the second quality state. When the medium quality state is characterized as the first quality state, the Nth processing layer is determined as the target partitioning layer, and when the medium quality state is characterized as the second quality state, the Mth processing layer is determined as the target partitioning layer. N and M are both positive integers, and N is greater than M.

[0114] In actual applications, if the end-cloud integrated detection model has 10 processing layers, when N is 8, it corresponds to the end-cloud integrated detection model A, which includes the end-side running sub-model A and the cloud-side running sub-model A. When M is 2, it corresponds to the end-cloud integrated detection model B, which includes the end-side running sub-model B and the cloud-side running sub-model B. The end-side running sub-model A and the end-side running sub-model B are deployed on the terminal, and the cloud-side running sub-model A and the cloud-side running sub-model B are deployed on the cloud server.

[0115] When the intermediate quality state is characterized as the first quality state, the target image is input to the end-side running sub-model A, and after the intermediate layer features are output, the intermediate layer features are sent to the cloud server, and the cloud-side running sub-model A detects the intermediate layer features.

[0116] When the intermediate quality state is characterized as the second quality state, the target image is input to the end-side running sub-model B, and after the intermediate layer features are output, the intermediate layer features are sent to the cloud server, and the cloud-side running sub-model B detects the intermediate layer features.

[0117] Steps S401 to S402 shown in this embodiment further divide the medium quality status. When the network status is poor, the terminal runs more links and transmits a small amount of information to the cloud server. When the network status is good, the terminal runs fewer links and transmits a large amount of information to the cloud server, further improving the detection accuracy while ensuring the detection speed.

[0118] See also Figure 5 In some embodiments, step S302 may include but is not limited to steps S501 to S502:

[0119] Step S501, compressing the intermediate layer features through a preset coding and compression module to obtain intermediate layer compressed features; wherein the intelligent compression model includes a coding and compression module and a decoding module, the coding and compression module is deployed on the terminal, and the decoding module is deployed on the cloud server;

[0120] In step S502, the intermediate layer compressed features are sent to the cloud server, and the cloud server decodes the intermediate layer compressed features through a decoding module to obtain intermediate layer decoded features, and performs object detection on the intermediate layer decoded features through a cloud-running sub-model to obtain a third detection result.

[0121] In step S501 of some embodiments, after the terminal side runs the sub-model to output the intermediate layer features, the intermediate layer features need to be sent to the cloud server. In order to reduce the amount of data that needs to be sent, the intelligent compression model can be pre-trained to decompress the data. The intelligent compression model includes an encoding and compression module and a decoding module. The encoding and compression module is deployed on the terminal, and the decoding module is deployed on the cloud server.

[0122] Specifically, after the terminal side runs the sub-model to output the intermediate layer features, the intermediate layer features are compressed through the encoding compression module to obtain the intermediate layer compressed features.

[0123] In step S502 of some embodiments, the intermediate layer compressed features are sent to the cloud server, and the cloud server decodes the intermediate layer compressed features through a decoding module to obtain intermediate layer decoded features. The intermediate layer decoded features are input into the cloud running sub-model to output the third detection result.

[0124] Steps S501 to S502 shown in this embodiment reduce the amount of data transmitted from the terminal to the cloud server, are more adaptable to poor network conditions, and improve transmission speed.

[0125] See also Figure 6 In some embodiments, before step S501, the image detection method based on network state adaptive perception further includes training the end-cloud integrated detection model, and the training process includes but is not limited to steps S601 to S607:

[0126] Step S601, obtaining an image sample;

[0127] Step S602, input the image sample to the end side and run the sub-model to obtain the intermediate features of the sample;

[0128] Step S603, inputting the sample intermediate features into the encoding compression module to obtain the sample intermediate compressed features;

[0129] Step S604, inputting the sample intermediate compression feature into a decoding module to obtain the sample intermediate decoding feature;

[0130] Step S605, input the sample intermediate decoding features into the cloud running sub-model to obtain the sample detection results;

[0131] Step S606, calculating loss data according to the sample detection result and the preset detection label to obtain target loss data;

[0132] Step S607: Adjust parameters of the end-cloud integrated detection model according to the target loss data to obtain a trained end-cloud integrated detection model.

[0133] In steps S601 to S607 shown in this embodiment, after the end-cloud integrated detection model and the intelligent compression model are pre-trained, the end-cloud integrated detection model and the intelligent compression model are jointly trained to reduce the loss caused by the reconstruction error, because the actual input to the cloud operation sub-model is the intermediate layer compression feature, not the intermediate layer feature. In this joint training link, the parameters of the intelligent compression module remain unchanged.

[0134] See also Figure 7 In some embodiments, before step S101, the image detection method based on network state adaptive perception may further include constructing a first detection model, and the construction process includes but is not limited to steps S701 to S706:

[0135] Step S701, obtaining an intelligent pruning model; wherein the intelligent pruning model includes an initial detection model and a parameter importance evaluation model, the parameter importance evaluation model is used to evaluate the importance of parameters of each processing layer in the initial detection model, and the initial detection model includes multiple processing layers;

[0136] Step S702, obtaining training image data;

[0137] Step S703, inputting the training image data into the initial detection model for object detection to obtain output features of each processing layer;

[0138] Step S704, inputting the parameters of each processing layer and each output feature in the initial detection model into the parameter importance evaluation model for importance evaluation to obtain the importance score of each parameter;

[0139] Step S705, determining the parameters whose importance scores are lower than the preset threshold as target parameters, and setting the target parameters in the initial detection model to zero;

[0140] Step S706: Adjust parameters of the initial detection model based on a preset loss function to obtain a first detection model.

[0141] In step S701 of some embodiments, the initial detection model is a lightweight detection model that has been pre-trained in a conventional manner. The lightweight detection model still has room for further compression and optimization, so intelligent pruning is performed after the training is completed.

[0142] Specifically, the intelligent pruning model includes an initial detection model and a parameter importance evaluation model, and the parameter importance evaluation model is used to evaluate the importance of parameters of each processing layer in the initial detection model. The initial detection model includes multiple processing layers;

[0143] In steps S702 to S704 of some embodiments, the loss function of the intelligent pruning model includes regression loss, classification loss and importance score minimization loss. Among them, the importance score minimization loss is used to make the sum of the importance scores of all parameters as small as possible. The training image data is input into the initial detection model for object detection to obtain the output features of each processing layer. The parameters of each processing layer in the initial detection model and the output features of each processing layer are input into the parameter importance evaluation model for importance evaluation to obtain the importance score of each parameter.

[0144] In some embodiments, in steps S705 to S706, the parameters whose importance scores are lower than a preset threshold are determined as target parameters, and the target parameters in the initial detection model are set to zero. The parameters of the initial detection model are adjusted based on the loss function until the loss function converges to obtain the first detection model.

[0145] In practical applications, the training method may be: first, the initial detection model is trained for 10 rounds, and the parameters of the initial detection model are fixed. Then, the parameter importance evaluation model is trained for 10 rounds, and the importance score of each parameter is output according to the parameter importance evaluation model, and the target parameters whose importance scores in the initial detection model are lower than the preset threshold are pruned, that is, the target parameters are set to zero. The above steps are repeated until the loss function converges to obtain the first detection model.

[0146] In steps S701 to S706 shown in this embodiment, by intelligently pruning the initial detection model, on the one hand, the computational complexity of the detection model is reduced, and on the other hand, the performance of the detection model is ensured to remain unchanged or slightly improved.

[0147] See also Figure 8 In some embodiments, step S102 includes but is not limited to steps S801 to S804:

[0148] Step S801, obtaining interaction data between the terminal and the cloud server;

[0149] Step S802, parsing the interaction data to obtain a plurality of first data indicators;

[0150] Step S803, inputting the plurality of first data indicators into a preset network status prediction model to perform network status prediction, and obtaining a plurality of second data indicators within a preset time period;

[0151] Step S804: obtaining a network status according to the plurality of second data indicators and a network status mapping table, wherein the network status mapping table contains network statuses corresponding to second data indicators of different combinations.

[0152] In steps S801 to S802 of some embodiments, the terminal analyzes key data indicators by performing random data interaction with the target server. Multiple first data indicators include: network bandwidth (unit: bps), network speed (unit: bps), network availability (unit: %), packet loss rate (unit: %), retransmission rate (unit: %) and connection establishment time (unit: ms). In some embodiments, the average value of the above data indicators is calculated in units of 1s as the first data indicator. Among them, the duration of 1s can be adjusted according to the actual business needs.

[0153] In steps S803 to S804 of some embodiments, multiple first data indicators are input into a preset network state prediction model to predict the network state, and multiple second data indicators within a preset time period are obtained. Different combinations of multiple second data indicators are mapped into different network states to obtain a network state mapping table. According to the multiple second data indicators and the network state mapping table, the network state can be obtained.

[0154] Steps S801 to S804 shown in this embodiment can predict the network status between the terminal and the cloud server within a preset time period, so as to select different detection methods for detection according to the network status.

[0155] See also Fig. 9 The embodiment of the present application further provides an image detection device based on adaptive network state perception, which can implement the above-mentioned image detection method based on adaptive network state perception, and the device includes:

[0156] An acquisition module 901 is used to acquire a target image;

[0157] Prediction module 902, used to predict the network status between the terminal and the cloud server within a preset time period; wherein the network status includes a low quality status and a high quality status;

[0158] The detection module 903 is used to perform object detection on the target image through a preset first detection model to obtain a first detection result if the network state is characterized as a low-quality state; wherein the first detection model is a detection model deployed on the terminal;

[0159] A sending module 904 is used to send the target image to the cloud server if the network state is characterized as a high-quality state, and the cloud server performs object detection on the target image through a preset second detection model to obtain a second detection result; wherein the number of processing layers included in the second detection model is greater than the number of processing layers included in the first detection model;

[0160] The receiving module 905 is used to receive the second detection result returned by the cloud server.

[0161] In some embodiments, the image detection device based on network state adaptive perception further includes a segmentation module. Before acquiring the target image, the segmentation module is used to:

[0162] Obtain an end-to-end detection model, which includes multiple processing layers.

[0163] determining a target partitioning layer among the plurality of processing layers;

[0164] Among the multiple processing layers, a terminal-side operation sub-model is formed according to the processing layers between the first processing layer and the target division layer, and the terminal-side operation sub-model is deployed on the terminal;

[0165] Among the multiple processing layers, a cloud-based operation sub-model is formed according to the processing layers from the processing layer after the target division layer to the last processing layer, and the cloud-based operation sub-model is deployed on a cloud server;

[0166] After obtaining the network status between the terminal and the cloud server, the image detection device based on network status adaptive perception also includes a terminal cloud detection module for:

[0167] If the network status is characterized as a medium quality state, the target image is input to the client-side running sub-model for object detection to obtain the intermediate layer features;

[0168] The intermediate layer features are sent to the cloud server, and the cloud server performs object detection on the intermediate layer features by running the sub-model in the cloud to obtain a third detection result;

[0169] Receive the third detection result returned by the cloud server.

[0170] In some embodiments, the image detection device based on adaptive perception of network status further includes a training module. Before the intermediate layer features are compressed by a preset coding compression module to obtain the intermediate layer compressed features, the training module is used to train the end-cloud integrated detection model, specifically including:

[0171] Get image samples;

[0172] Run the sub-model on the input side of the image sample to obtain the intermediate features of the sample;

[0173] Input the sample intermediate features into the encoding compression module to obtain the sample intermediate compressed features;

[0174] Input the sample intermediate compression features into the decoding module to obtain the sample intermediate decoding features;

[0175] Input the sample intermediate decoding features into the cloud-based running sub-model to obtain the sample detection results;

[0176] Calculate the loss data based on the sample test results and the preset test labels to obtain the target loss data;

[0177] The parameters of the end-cloud integrated detection model are adjusted according to the target loss data to obtain the trained end-cloud integrated detection model.

[0178] In some embodiments, the image detection device based on network state adaptive perception further includes a pruning module. Before acquiring the target image, the pruning module is used to construct a first detection model, which specifically includes:

[0179] Obtaining an intelligent pruning model; wherein the intelligent pruning model includes an initial detection model and a parameter importance evaluation model, the parameter importance evaluation model is used to evaluate the importance of parameters of each processing layer in the initial detection model, and the initial detection model includes multiple processing layers;

[0180] Obtain training image data;

[0181] Input the training image data into the initial detection model for object detection and obtain the output features of each processing layer;

[0182] Input the parameters of each processing layer and each output feature in the initial detection model into the parameter importance evaluation model for importance evaluation to obtain the importance score of each parameter;

[0183] Determine the parameters whose importance scores are lower than a preset threshold as target parameters, and set the target parameters in the initial detection model to zero;

[0184] The parameters of the initial detection model are adjusted based on a preset loss function to obtain a first detection model.

[0185] The specific implementation of the image detection device based on adaptive network status perception is basically the same as the specific implementation of the image detection method based on adaptive network status perception described above, and will not be repeated here.

[0186] The embodiment of the present application also provides an electronic device, the electronic device includes a memory and a processor, the memory stores a computer program, and the processor implements the above-mentioned image detection method based on adaptive perception of network status when executing the computer program. The electronic device can be any intelligent terminal including a tablet computer, a car computer, etc.

[0187] See also Fig.10 , Fig.10 The hardware structure of an electronic device of another embodiment is illustrated, and the electronic device includes:

[0188] The processor 1001 may be implemented by a general-purpose CPU (Central Processing Unit), a microprocessor, an application-specific integrated circuit (ASIC), or one or more integrated circuits, and is used to execute relevant programs to implement the technical solutions provided in the embodiments of the present application.

[0189] The memory 1002 can be implemented in the form of a read-only memory (ROM), a static storage device, a dynamic storage device, or a random access memory (RAM). The memory 1002 can store an operating system and other applications. When the technical solution provided in the embodiment of this specification is implemented by software or firmware, the relevant program code is stored in the memory 1002, and the processor 1001 calls and executes the image detection method based on network state adaptive perception in the embodiment of this application;

[0190] Input / output interface 1003, used to implement information input and output;

[0191] The communication interface 1004 is used to realize the communication interaction between the device and other devices. The communication can be realized through a wired manner (such as USB, network cable, etc.) or a wireless manner (such as mobile network, WI FI, Bluetooth, etc.);

[0192] A bus 1005 , which transmits information between various components of the device (e.g., the processor 1001 , the memory 1002 , the input / output interface 1003 , and the communication interface 1004 );

[0193] The processor 1001 , the memory 1002 , the input / output interface 1003 and the communication interface 1004 are connected to each other in communication within the device via the bus 1005 .

[0194] An embodiment of the present application also provides a computer-readable storage medium, which stores a computer program. When the computer program is executed by a processor, the above-mentioned image detection method based on adaptive perception of network status is implemented.

[0195] The memory, as a non-transient computer-readable storage medium, can be used to store non-transient software programs and non-transient computer executable programs. In addition, the memory may include a high-speed random access memory, and may also include a non-transient memory, such as at least one disk storage device, a flash memory device, or other non-transient solid-state storage device. In some embodiments, the memory may optionally include a memory remotely disposed relative to the processor, and these remote memories may be connected to the processor via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0196] The embodiments described in the embodiments of the present application are intended to more clearly illustrate the technical solutions of the embodiments of the present application and do not constitute a limitation on the technical solutions provided in the embodiments of the present application. Those skilled in the art will appreciate that with the evolution of technology and the emergence of new application scenarios, the technical solutions provided in the embodiments of the present application are also applicable to similar technical problems.

[0197] Those skilled in the art will appreciate that the technical solutions shown in the figures do not constitute a limitation on the embodiments of the present application, and may include more or fewer steps than shown in the figures, or a combination of certain steps, or different steps.

[0198] The device embodiments described above are merely illustrative, and the units described as separate components may or may not be physically separated, that is, they may be located in one place or distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0199] Those skilled in the art will appreciate that all or some of the steps in the methods disclosed above, and the functional modules / units in the systems and devices may be implemented as software, firmware, hardware, or a suitable combination thereof.

[0200] The terms "first", "second", "third", "fourth", etc. (if any) in the specification of the present application and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments of the present application described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device comprising a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0201] It should be understood that in the present application, "at least one (item)" means one or more, and "plurality" means two or more. "And / or" is used to describe the association relationship of associated objects, indicating that three relationships may exist. For example, "A and / or B" can mean: only A exists, only B exists, and A and B exist at the same time, where A and B can be singular or plural. The character " / " generally indicates that the objects associated before and after are in an "or" relationship. "At least one of the following" or similar expressions refers to any combination of these items, including any combination of single or plural items. For example, at least one of a, b or c can mean: a, b, c, "a and b", "a and c", "b and c", or "a and b and c", where a, b, c can be single or multiple.

[0202] In the several embodiments provided in the present application, it should be understood that the disclosed devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of the above units is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0203] The units described above as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0204] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0205] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application is essentially or the part that contributes to the prior art or all or part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including multiple instructions to enable a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of various embodiments of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (Read-Only Memory, referred to as ROM), random access memory (Random Access Memory, referred to as RAM), disk or optical disk and other media that can store programs.

[0206] The preferred embodiments of the present invention are described above with reference to the accompanying drawings, but the scope of the rights of the present invention is not limited thereto. Any modification, equivalent substitution and improvement made by a person skilled in the art without departing from the scope and essence of the present invention should be within the scope of the rights of the present invention.

Claims

1. An image detection method based on adaptive perception of network status, characterized in that: Applied to a terminal, the method comprises: Get the target image; Predicting the network state between the terminal and the cloud server within a preset time period; wherein the network state includes a low-quality state, a medium-quality state, and a high-quality state, the medium-quality state includes a first quality state and a second quality state, and the network state corresponding to the first quality state is worse than the network state corresponding to the second quality state; Acquire an end-cloud integrated detection model, wherein the end-cloud integrated detection model includes multiple processing layers; Determining a target partitioning layer among the plurality of processing layers; wherein determining a target partitioning layer among the plurality of processing layers comprises: If the medium quality state is characterized as the first quality state, then, among the multiple processing layers, determining the Nth processing layer as the target partitioning layer; If the medium quality state is characterized as the second quality state, then among the multiple processing layers, the Mth processing layer is determined as the target partitioning layer; wherein N and M are both positive integers, and N is greater than M; Among the multiple processing layers, a terminal-side operation sub-model is formed according to the processing layers between the first processing layer and the target division layer, and the terminal-side operation sub-model is deployed on the terminal; Among the plurality of processing layers, a cloud operation sub-model is formed according to the processing layers from the processing layer after the target division layer to the last processing layer, and the cloud operation sub-model is deployed on the cloud server; If the network state is characterized as a low-quality state, object detection is performed on the target image through a preset first detection model to obtain a first detection result; wherein the first detection model is a detection model deployed on the terminal; If the network state is characterized as a high-quality state, the target image is sent to the cloud server, and the cloud server performs object detection on the target image through a preset second detection model to obtain a second detection result; wherein the number of processing layers included in the second detection model is greater than the number of processing layers included in the first detection model; Receiving the second detection result returned by the cloud server; If the network state is characterized as a medium quality state, the target image is input into the end-side running sub-model for object detection to obtain intermediate layer features; The intermediate layer features are sent to the cloud server, and the cloud server performs object detection on the intermediate layer features through the cloud operation sub-model to obtain a third detection result; Receive the third detection result returned by the cloud server.

2. The method according to claim 1, characterized in that The sending of the intermediate layer features to the cloud server, and the cloud server performing object detection on the intermediate layer features through the cloud running sub-model to obtain a third detection result, includes: The intermediate layer features are compressed by a preset coding and compression module to obtain intermediate layer compressed features; wherein the intelligent compression model includes a coding and compression module and a decoding module, the coding and compression module is deployed on the terminal, and the decoding module is deployed on the cloud server; The intermediate layer compression features are sent to the cloud server, and the cloud server decodes the intermediate layer compression features through the decoding module to obtain intermediate layer decoded features, and performs object detection on the intermediate layer decoded features through the cloud running sub-model to obtain the third detection result.

3. The method according to claim 2, characterized in that Before compressing the intermediate layer features by a preset encoding compression module to obtain the intermediate layer compressed features, the method further includes: The end-cloud integrated detection model is trained, including: Get image samples; Inputting the image sample into the end-side running sub-model to obtain intermediate features of the sample; Inputting the sample intermediate features into the encoding compression module to obtain sample intermediate compressed features; Inputting the sample intermediate compression feature into the decoding module to obtain the sample intermediate decoding feature; Inputting the sample intermediate decoding features into the cloud-based running sub-model to obtain a sample detection result; Calculating loss data according to the sample detection results and preset detection labels to obtain target loss data; Parameters of the end-cloud integrated detection model are adjusted according to the target loss data to obtain the trained end-cloud integrated detection model.

4. The method according to claim 1, characterized in that: Before acquiring the target image, the method further includes: Constructing the first detection model includes: Acquire an intelligent pruning model; wherein the intelligent pruning model includes an initial detection model and a parameter importance evaluation model, the parameter importance evaluation model is used to evaluate the importance of parameters of each processing layer in the initial detection model, and the initial detection model includes multiple processing layers; Obtain training image data; Inputting the training image data into the initial detection model to perform object detection, and obtaining output features of each processing layer; Inputting the parameters of each processing layer in the initial detection model and each output feature into the parameter importance evaluation model for importance evaluation to obtain the importance score of each parameter; Determine the parameter whose importance score is lower than a preset threshold as the target parameter, and set the target parameter in the initial detection model to zero; The parameters of the initial detection model are adjusted based on a preset loss function to obtain the first detection model.

5. The method according to claim 1, characterized in that The predicting of the network status between the terminal and the cloud server within a preset time period includes: Acquiring interaction data between the terminal and the cloud server; Parsing the interaction data to obtain a plurality of first data indicators; Inputting the plurality of the first data indicators into a preset network status prediction model to perform network status prediction, and obtaining a plurality of second data indicators within a preset time period; The network status is obtained according to a plurality of the second data indicators and a network status mapping table, wherein the network status mapping table contains network statuses corresponding to different combinations of second data indicators.

6. An image detection device based on adaptive perception of network status, characterized in that: The device comprises: An acquisition module, used for acquiring a target image; A prediction module, used to predict the network state between the terminal and the cloud server within a preset time period; wherein the network state includes a low quality state, a medium quality state and a high quality state, the medium quality state includes a first quality state and a second quality state, and the network state corresponding to the first quality state is worse than the network state corresponding to the second quality state; Acquire an end-cloud integrated detection model, wherein the end-cloud integrated detection model includes multiple processing layers; Determining a target partitioning layer among the plurality of processing layers; wherein determining a target partitioning layer among the plurality of processing layers comprises: If the medium quality state is characterized as the first quality state, then, among the multiple processing layers, determining the Nth processing layer as the target partitioning layer; If the medium quality state is characterized as the second quality state, then among the multiple processing layers, the Mth processing layer is determined as the target partitioning layer; wherein N and M are both positive integers, and N is greater than M; Among the multiple processing layers, a terminal-side operation sub-model is formed according to the processing layers between the first processing layer and the target division layer, and the terminal-side operation sub-model is deployed on the terminal; Among the plurality of processing layers, a cloud operation sub-model is formed according to the processing layers from the processing layer after the target division layer to the last processing layer, and the cloud operation sub-model is deployed on the cloud server; A detection module, configured to perform object detection on the target image through a preset first detection model to obtain a first detection result if the network state is characterized as a low-quality state; wherein the first detection model is a detection model deployed on the terminal; a sending module, configured to send the target image to the cloud server if the network state is characterized as a high-quality state, and the cloud server performs object detection on the target image through a preset second detection model to obtain a second detection result; wherein the number of processing layers included in the second detection model is greater than the number of processing layers included in the first detection model; If the network state is characterized as a medium quality state, the target image is input into the end-side running sub-model for object detection to obtain intermediate layer features; The intermediate layer features are sent to the cloud server, and the cloud server performs object detection on the intermediate layer features through the cloud operation sub-model to obtain a third detection result; A receiving module is used to receive the second detection result returned by the cloud server, and receive the third detection result returned by the cloud server.

7. An electronic device, characterized in that: The electronic device includes a memory and a processor, the memory stores a computer program, and the processor implements the image detection method based on network status adaptive perception as described in any one of claims 1 to 5 when executing the computer program.

8. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the image detection method based on adaptive perception of network status according to any one of claims 1 to 5 is implemented.

Images