Image detection method, device and storage medium

Abstract

Embodiments of the present application provide an image detection method and device and a storage medium, relating to the technical field of image detection, which comprises: generating a plurality of initial small map images by using a blank small map image and a to-be-detected object image; performing feature enhancement on the plurality of initial small map images according to game image features to obtain a training data set, training an initial detection network of an image detection model by using the training data set to obtain a target detection network; extracting a small map region from a to-be-detected image by using a small map detection network of the image detection model, and obtaining a to-be-detected small map image from the small map region according to small map features; and performing target detection on the to-be-detected small map image by using the target detection network to obtain target information. The present application can accurately locate a small map from a game image and quickly detect target information from the small map, thereby providing real-time data support for game event analysis and tactical decision-making.

Description

Technical Field

[0001] This invention relates to the field of image detection technology, and more specifically, to an image detection method, device, and storage medium. Background Technology

[0002] In multiplayer online competitive games, a minimap is typically displayed on the game screen to help players obtain crucial information about the overall game situation. This minimap usually records key tactical elements such as friendly and enemy characters, as well as various critical resources. With the development of esports competitions, the demand for match analysis and tactical replays is constantly increasing. Therefore, accurately and quickly extracting key information from the minimap from the game screen has become a pressing technical problem that needs to be solved. Summary of the Invention

[0003] In view of this, the purpose of the present invention is to provide an image detection method, device and storage medium to quickly extract key information from the minimap in the game interface.

[0004] To achieve the above objectives, the technical solutions adopted in the embodiments of the present invention are as follows: In a first aspect, embodiments of the present invention provide an image detection method, comprising: Multiple initial minimap images are generated using a blank minimap image and an image of the object to be detected; Based on the features of the competition images, feature enhancement is performed on multiple initial small map images to obtain a training dataset. The initial detection network of the image detection model is then trained using the training dataset to obtain the target detection network. The image detection model uses a small map detection network to extract small map regions from the image to be detected, and obtains the small map image to be tested from the small map regions based on the small map features. The target information is obtained by performing target detection on the small map image to be tested using the target detection network.

[0005] In an optional implementation, the step of generating multiple initial minimap images using the blank minimap image and the image of the object to be detected includes: The image of the object to be detected is cropped according to the features of the competition image to obtain a cropped image, and a corresponding outer border is assigned to the cropped image to obtain the competition object image; Based on the characteristics of the competition image, the image of the competition object is added to the blank minimap image to obtain multiple initial minimap images.

[0006] In an optional implementation, the step of performing feature enhancement on the multiple initial minimap images based on the features of the competition image includes: The initial minimap image is enhanced based on the perturbation of the outer border color, the perturbation of the size of the competition object image, and the perturbation of the blur of the competition object image; and / or The initial minimap image is enhanced by compressing the image corresponding to the competition image, overlapping the images of the competition objects, and simulating the paths between the competition objects.

[0007] In an optional implementation, the step of training the initial detection network of the image detection model using the training dataset to obtain the target detection network includes: Based on the category and style of the image of the object to be detected, the training dataset is fused at multiple scales to obtain a fused dataset; The initial detection network is trained using the fused dataset, and the target detection network is obtained by supervising the loss of the initial detection network through a joint loss function.

[0008] In an optional implementation, the step of constructing the joint loss function includes: Assign a detection label to each sample in the fused dataset; Using the detection labels, calculate the bounding box regression loss, binary cross-entropy classification loss, and distributed focus loss of the initial detection network, respectively. The joint loss function is constructed based on the bounding box regression loss, the binary cross-entropy classification loss, and the distributed focus loss.

[0009] In an optional implementation, the step of obtaining target information by performing target detection on the small map image to be tested through the target detection network includes: The object detection network extracts the class probability and bounding box parameters of each object image from the small map image under test; The object category of each object image is determined based on the category probability and the bounding box parameters, and the target information is obtained based on the object category.

[0010] In an optional implementation, the step of extracting the small map region from the image to be detected includes: Determine the perspective of the person corresponding to the image to be detected; If the character's perspective is a first-person perspective, then candidate regions are selected from the image to be detected; The small map region is extracted from the candidate region using the border color feature of the small map; If the character's perspective is a second perspective, then the minimap region is extracted from a fixed area of ​​the image to be detected.

[0011] Secondly, embodiments of the present invention provide an image detection device, comprising: The image generation module is used to generate multiple initial minimap images using a blank minimap image and an image of the object to be detected. The network training module is used to perform feature enhancement on multiple initial small map images based on the features of the competition images to obtain a training dataset, and to train the initial detection network of the image detection model using the training dataset to obtain the target detection network. The map extraction module is used to extract the small map region from the image to be detected through the small map detection network of the image detection model, and to obtain the small map image to be tested from the small map region according to the small map features; The target detection module is used to perform target detection on the small map image to be tested through the target detection network to obtain target information.

[0012] Thirdly, embodiments of the present invention provide an image detection device, including a processor and a memory, wherein the memory stores machine-executable instructions that can be executed by the processor, and the processor can execute the machine-executable instructions to implement the image detection method described in the first aspect.

[0013] Fourthly, embodiments of the present invention provide a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the image detection method described in the first aspect.

[0014] The image detection method provided in this invention generates multiple initial minimap images using a blank minimap image and an image of the object to be detected. Then, it performs feature enhancement processing on the initial minimap images based on the features of the game image to obtain a training dataset. The training dataset is used to train the target detection network of the image recognition model, enabling the image recognition model to accurately locate the minimap in the game image and quickly detect target information from the minimap, thereby providing real-time data support for game event analysis and tactical decision-making.

[0015] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, preferred embodiments are described below in detail with reference to the accompanying drawings. Attached Figure Description

[0016] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of the present invention and should not be regarded as a limitation on the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0017] Figure 1A block diagram of an image detection device provided in an embodiment of the present invention is shown; Figure 2 A flowchart illustrating an image detection method provided by an embodiment of the present invention is shown; Figure 3 A schematic diagram of a game interface provided by an embodiment of the present invention is shown; Figure 4 A functional block diagram of an image detection device provided by an embodiment of the present invention is shown.

[0018] icon: 100 - Image detection device; 110 - Memory; 120 - Processor; 130 - Communication module; 200 - Image generation module; 300 - Network training module; 400 - Map extraction module; 500 - Target detection module. Detailed Implementation

[0019] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. The components of the embodiments of the present invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations.

[0020] Therefore, the following detailed description of the embodiments of the invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely to illustrate selected embodiments of the invention. All other embodiments obtained by those skilled in the art based on the embodiments of the invention without inventive effort are within the scope of protection of the invention.

[0021] It should be noted that relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0022] Please refer to Figure 1This is a block diagram of an image detection device 100. The image detection device 100 includes a memory 110, a processor 120, and a communication module 130. The memory 110, processor 120, and communication module 130 are electrically connected to each other directly or indirectly to achieve data transmission or interaction. For example, these components can be electrically connected to each other through one or more communication buses or signal lines.

[0023] The memory 110 is used to store programs or data. The memory 110 may be, but is not limited to, random access memory (RAM), read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), etc.

[0024] The processor 120 is used to read / write data or programs stored in the memory 110 and to perform corresponding functions.

[0025] The communication module 130 is used to establish a communication connection between the image detection device 100 and other communication terminals through the network, and to send and receive data through the network.

[0026] It should be understood that, Figure 1 The structure shown is only a schematic diagram of the image detection device 100. The image detection device 100 may also include a larger... Figure 1 The more or fewer components shown, or having the same Figure 1 The different configurations shown. Figure 1 The components shown can be implemented using hardware, software, or a combination thereof.

[0027] Please refer to Figure 2 The figure is a schematic flowchart of an image detection method provided in this embodiment, which includes: S201. Generate multiple initial minimap images using blank minimap images and images of objects to be detected.

[0028] Please refer to Figure 3 , Figure 3 This is a schematic diagram of a game interface provided in this embodiment.

[0029] exist Figure 3In this diagram, A represents the game interface, B represents the minimap, C represents the respawn time of important resources, D represents important resources, and E represents game characters. Within a single game interface, there can be multiple game characters (E), important resources (D), and important resource respawn times (C). Since match analysis, tactical replays, and strategic planning typically revolve around information such as game characters (E), important resources (D), and important resource respawn times (C), this information from the minimap can be used as the target for analysis.

[0030] In some popular multiplayer competitive games, there are usually multiple game characters, but the number of game characters appearing in a match is limited. Therefore, obtaining relevant images of the minimap from match videos or other sources may result in the inability to obtain most of the information related to the game characters, affecting the recognition performance of the image detection model.

[0031] Therefore, game-related information can be obtained through some dedicated interfaces or methods. For example, images of all objects to be detected, as well as various blank minimap images, can be downloaded from the game's official website. Blank minimap images can be minimap images that do not contain any objects to be detected, or minimap images of the game at certain specific moments, such as the start or end of the game.

[0032] After obtaining the blank minimap image and the image of the object to be detected, the minimap image in the actual game can be simulated using these images. Since in actual gameplay, the objects to be detected, such as game characters, may be distributed across various locations on the minimap, multiple initial minimap images can be obtained by following the game rules, for example, randomly selecting 10 characters from all game characters, randomly assigning 5 characters to each side, and then randomly distributing these characters across various locations on the minimap.

[0033] S202. Based on the features of the competition images, feature enhancement is performed on multiple initial small map images to obtain a training dataset. The initial detection network of the image detection model is trained using the training dataset to obtain the target detection network.

[0034] Since esports competitions are typically held online, game visuals are affected by various factors, such as network connectivity, device hardware and software configurations, which can lead to blurry, ghosting, or flickering images. Therefore, to improve the model's detection capabilities, data augmentation can be performed on the initial smallmap image to obtain a training dataset based on the potential issues encountered in esports competitions. This training dataset can then be used to train the initial detection network of the image detection model, resulting in the object detection network.

[0035] S203. Extract the small map region from the image to be detected using the small map detection network of the image detection model, and obtain the small map image to be tested from the small map region based on the small map features.

[0036] The image to be tested can be a game screen from an e-sports competition. The image detection model can acquire the game screen in real time, determine the approximate small map area from the game screen through the small map detection network, and then obtain the small map image to be tested from the small map area.

[0037] S204. Target information is obtained by performing target detection on the small map image to be tested through the target detection network.

[0038] The trained object detection network is used to detect objects in the small map image to be tested, so as to identify the target objects in the small map. Then, based on the information of the target objects in each small map, the corresponding target information is generated.

[0039] This embodiment generates multiple initial minimap images using a blank minimap image and an image of the object to be detected. Then, based on the features of the game image, the initial minimap images are subjected to feature enhancement processing to obtain a training dataset. The target detection network of the image recognition model is trained using the training dataset, enabling the image recognition model to accurately locate the minimap in the game image and quickly detect target information from the minimap, thereby providing real-time data support for game event analysis and tactical decision-making.

[0040] In one embodiment, the step of generating multiple initial minimap images using the blank minimap image and the image of the object to be detected includes: The image of the object to be detected is cropped according to the features of the competition image to obtain a cropped image, and a corresponding outer border is assigned to the cropped image to obtain the competition object image; Based on the characteristics of the competition image, the image of the competition object is added to the blank minimap image to obtain multiple initial minimap images.

[0041] In esports game visuals, the size of the images of objects to be detected on the minimap is usually fixed or set according to fixed parameters. Therefore, to eliminate potential interference from other non-object images, the acquired object images can be cropped or scaled down accordingly to ensure that the size of each object image is consistent with the size in the esports game. Then, based on the image characteristics in the esports game, a corresponding outer border is assigned to the cropped image.

[0042] For example, in regular esports competitions, multiple professional players are usually divided into red and blue teams. During the match, the red team's game characters are given a red border on the minimap, while the blue team's game characters are given a blue border on the minimap.

[0043] At the start of an esports match, the red team's area contains only red team characters, and the blue team's area contains only blue team characters. As the match progresses, the positions of the characters may change, at which point blue team characters may be detected in the red team's area. Therefore, based on the match's progress or time, images of the match objects can be added to a blank minimap image to obtain multiple initial minimap images.

[0044] This embodiment uses image features from actual esports matches to add images of the competition objects to a blank minimap image to obtain multiple initial minimap images, providing data support for subsequent model training. This allows the trained model to accurately extract the desired information from the competition images at any time.

[0045] In one implementation, the step of extracting the minimap region from the image to be detected includes: Determine the perspective of the person corresponding to the image to be detected; If the character's perspective is a first-person perspective, then candidate regions are selected from the image to be detected; The small map region is extracted from the candidate region using the border color feature of the small map; If the character's perspective is a second perspective, then the minimap region is extracted from a fixed area of ​​the image to be detected.

[0046] The first-person perspective is the player's view, and the second-person perspective is the match view. Since the minimap's position is usually fixed in the match view, the minimap area can be directly extracted from this fixed area. However, players may adjust the minimap's position according to their habits, so candidate areas can be determined based on common player settings, and then the minimap area can be extracted from these candidate areas.

[0047] After determining the minimap area, you can use color filtering to locate the minimap.

[0048]

[0049] Where H, S, and V represent the HSV values ​​of pixels after the image is converted to HSV format, min and max represent the filtering thresholds, and g and d represent two different colors.

[0050] For example, a gold border can be given to the minimap, where 'g' represents gold and 'd' represents other colors. The minimap can be located by filtering by color.

[0051] Then, the outer contour C, which includes the inner circle, is determined through contour detection:

[0052] in, - The outer contour contained in the image.

[0053] For each contour, calculate its minimum bounding rectangle. :

[0054] in, for A function used to calculate the minimum bounding matrix of the outer contour. This represents the coordinates of the top-left corner of the circumscribed matrix. , This represents the height of the circumscribed matrix.

[0055] Filter rectangles that meet the aspect ratio and position criteria:

[0056] in, This is the proportionality coefficient. This refers to the height of the game screen.

[0057] The largest valid rectangle was ultimately selected as the minimap area. :

[0058] in, This indicates taking the maximum value.

[0059] This embodiment selects different small map detection methods from different perspectives, and then extracts the small map region through filtering and circumscribed matrix algorithms, making the detection and extraction of small maps more accurate.

[0060] In one implementation, the step of performing feature enhancement on the plurality of initial small map images based on the features of the competition image includes: The initial minimap image is enhanced based on the perturbation of the outer border color, the perturbation of the size of the competition object image, and the perturbation of the blur of the competition object image; and / or The initial minimap image is enhanced by compressing the image corresponding to the competition image, overlapping the images of the competition objects, and simulating the paths between the competition objects.

[0061] Because game visuals can be affected by various factors during esports competitions, feature enhancements can be applied to the initial minimap image based on situations that have occurred or may occur during the competition.

[0062] Feature enhancement methods can simulate potential interference in the game screen, such as: outer ring color perturbation, avatar size perturbation, countdown color perturbation, blurring (through random scaling and restoration), image compression, simulating situations where only part of the character's avatar is visible in certain scenes, and connecting the character's avatar with a thin white line to simulate path guidance.

[0063] This embodiment enhances the initial small map image based on the possible interference in e-sports competitions, so that the training dataset can encompass small map images under various conditions, thereby improving the stability and reliability of model detection.

[0064] In one implementation, the step of training the initial detection network of the image detection model using the training dataset to obtain the target detection network includes: Based on the category and style of the image of the object to be detected, the training dataset is fused at multiple scales to obtain a fused dataset; The initial detection network is trained using the fused dataset, and the target detection network is obtained by supervising the loss of the initial detection network through a joint loss function.

[0065] like Figure 3 As shown, since there may be multiple important resources D, and each important resource D may have a different appearance, the style of the important resource refresh time C configured for each important resource D may also be different in order to distinguish the refresh time of each important resource D. Furthermore, different game characters correspond to different images.

[0066] Therefore, before training the initial detection network, the training set can be multi-scale fused according to the category and style of each important resource or each game character to obtain a fused training set, and then the initial detection network can be trained using the fused training set.

[0067] In this embodiment, when training the object detection network, multi-scale fusion is performed on the images in the training set, which enables the object detection network to accurately detect each object, thereby improving the detection performance and robustness of the object detection model.

[0068] In one implementation, the step of constructing the joint loss function includes: Assign a detection label to each sample in the fused dataset; Using the detection labels, calculate the bounding box regression loss, binary cross-entropy classification loss, and distributed focus loss of the initial detection network, respectively. The joint loss function is constructed based on the bounding box regression loss, the binary cross-entropy classification loss, and the distributed focus loss.

[0069] The joint loss function L is:

[0070] This represents the bounding box regression loss. This represents the binary cross-entropy classification loss. Represents distributed focus loss. , and All are weighting coefficients.

[0071] In one embodiment, the step of obtaining target information by performing target detection on the small map image to be tested through the target detection network includes: The object detection network extracts the class probability and bounding box parameters of each object image from the small map image under test; The object category of each object image is determined based on the category probability and the bounding box parameters, and the target information is obtained based on the object category.

[0072] The target detection network can adopt an anchor-free target detection framework, which can effectively reduce anchor-box matching conflicts and improve positioning accuracy when multiple targets are dense and heavily overlapped.

[0073] The bounding box parameters and class probabilities of each target object are extracted from the fused image using an object detection network. The bounding box parameters are the position and size of the border of the target object image, and the class probabilities are the probability of the target object image belonging to that class.

[0074] Since multiple bounding boxes may exist in a target image during target detection, methods such as maximum suppression can be used to remove duplicate bounding boxes based on the class probability corresponding to each bounding box and determine the target object corresponding to the target image. Then, the target information can be determined based on the information of the target object in each minimap image.

[0075] For example, if the probability of a target object image belonging to category A is determined to be 70% and the probability of it belonging to category B is 30%, then the target object image can be identified as A. Then, by obtaining information related to A, the target information can be obtained.

[0076] This embodiment classifies and filters each image object in the minimap image by using category probability and bounding box parameters, thereby determining each target object contained in the minimap image. Then, by combining the information of target objects in all minimap images, the target information can be obtained, which effectively supports the image detection model's understanding of the game and provides users with better real-time strategies.

[0077] To perform the corresponding steps in the above embodiments and various possible methods, an implementation of an image detection device is given below. Please refer to [link / reference]. Figure 4 , Figure 4 This is a functional block diagram of an image detection device provided in an embodiment of the present invention. It should be noted that the basic principle and technical effects of the image detection device provided in this embodiment are the same as those in the above embodiments. For the sake of brevity, any parts not mentioned in this embodiment can be referred to the corresponding content in the above embodiments. The image detection device 100 includes: Image generation module 200 is used to generate multiple initial minimap images using a blank minimap image and an image of the object to be detected; The network training module 300 is used to perform feature enhancement on multiple initial small map images based on the features of the competition images to obtain a training dataset, and to train the initial detection network of the image detection model using the training dataset to obtain an object detection network; The map extraction module 400 is used to extract a small map region from the image to be detected through the small map detection network of the image detection model, and to obtain the small map image to be tested from the small map region according to the small map features. The target detection module 500 is used to perform target detection on the small map image to be tested through the target detection network to obtain target information.

[0078] Optionally, the above modules can be stored in the form of software or firmware. Figure 1 The memory shown is either stored in or embedded in the operating system (OS) of the image detection device, and can be used by... Figure 1 The processor executes the commands. Meanwhile, the data and program code required to execute these modules can be stored in memory.

[0079] In the several embodiments provided in this application, it should be understood that the disclosed apparatus and methods can also be implemented in other ways. The apparatus embodiments described above are merely illustrative; for example, the flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of apparatus, methods, and computer program products according to various embodiments of the present invention. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions marked in the blocks may occur in a different order than those marked in the drawings. For example, two consecutive blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in a block diagram and / or flowchart, and combinations of blocks in block diagrams and / or flowcharts, can be implemented using a dedicated hardware-based system that performs the specified function or action, or using a combination of dedicated hardware and computer instructions.

[0080] In addition, the functional modules in the various embodiments of the present invention can be integrated together to form an independent part, or each module can exist independently, or two or more modules can be integrated to form an independent part.

[0081] If the aforementioned functions are implemented as software functional modules and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this invention, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0082] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. An image detection method, characterized in that, include: Generating multiple initial minimap images using a blank minimap image and an image of the object to be detected includes: cropping the image of the object to be detected according to the features of the competition image to obtain a cropped image; assigning a corresponding outer border to the cropped image to obtain the competition object image; and adding the competition object image to the blank minimap image according to the features of the competition image to obtain multiple initial minimap images. Based on the features of the competition images, feature enhancement is performed on multiple initial small map images to obtain a training dataset. The initial detection network of the image detection model is then trained using the training dataset to obtain the target detection network. Extracting a small map region from an image to be detected using the small map detection network of the image detection model includes: determining the perspective of a person in the image to be detected; if the perspective is a first perspective, selecting a candidate region from the image to be detected; extracting the small map region from the candidate region using the border color features of the small map; if the perspective is a second perspective, extracting the small map region from a fixed region of the image to be detected; and obtaining the small map image to be tested from the small map region based on the small map features. The target information is obtained by performing target detection on the small map image to be tested using the target detection network.

2. The image detection method according to claim 1, characterized in that, The step of performing feature enhancement on multiple initial small map images based on the features of the competition image includes: The initial minimap image is enhanced based on the perturbation of the outer border color, the perturbation of the size of the competition object image, and the perturbation of the blur of the competition object image; and / or The initial minimap image is enhanced by compressing the image corresponding to the competition image, overlapping the images of the competition objects, and simulating the paths between the competition objects.

3. The image detection method according to claim 1, characterized in that, The step of training the initial detection network of the image detection model using the training dataset to obtain the target detection network includes: Based on the category and style of the image of the object to be detected, the training dataset is fused at multiple scales to obtain a fused dataset; The initial detection network is trained using the fused dataset, and the target detection network is obtained by supervising the loss of the initial detection network through a joint loss function.

4. The image detection method according to claim 3, characterized in that, The steps for constructing the joint loss function include: Assign a detection label to each sample in the fused dataset; Using the detection labels, calculate the bounding box regression loss, binary cross-entropy classification loss, and distributed focus loss of the initial detection network, respectively. The joint loss function is constructed based on the bounding box regression loss, the binary cross-entropy classification loss, and the distributed focus loss.

5. The image detection method according to any one of claims 1-4, characterized in that, The step of obtaining target information by performing target detection on the small map image to be tested through the target detection network includes: The object detection network extracts the class probability and bounding box parameters of each object image from the small map image under test; The object category of each object image is determined based on the category probability and the bounding box parameters, and the target information is obtained based on the object category.

6. An image detection device, characterized in that, include: An image generation module is used to generate multiple initial minimap images using a blank minimap image and an image of the object to be detected, including: cropping the image of the object to be detected according to the features of the competition image to obtain a cropped image; assigning a corresponding outer border to the cropped image to obtain a competition object image; and adding the competition object image to the blank minimap image according to the features of the competition image to obtain multiple initial minimap images. The network training module is used to perform feature enhancement on multiple initial small map images based on the features of the competition images to obtain a training dataset, and to train the initial detection network of the image detection model using the training dataset to obtain the target detection network. The map extraction module is used to extract a small map region from an image to be detected using the small map detection network of the image detection model. The extraction process includes: determining the viewpoint of the person in the image to be detected; if the viewpoint is a first viewpoint, selecting a candidate region from the image to be detected; extracting the small map region from the candidate region using the border color features of the small map; if the viewpoint is a second viewpoint, extracting the small map region from a fixed region of the image to be detected; and obtaining the small map image to be tested from the small map region based on the small map features. The target detection module is used to perform target detection on the small map image to be tested through the target detection network to obtain target information.

7. An image detection device, characterized in that, The method includes a processor and a memory, the memory storing machine-executable instructions that can be executed by the processor to implement the image detection method according to any one of claims 1-5.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the image detection method as described in any one of claims 1-5.

Images