Question sheet paging and target detection model training method and device, equipment and medium

Abstract

The embodiment of the invention provides a question sheet paging method, device, equipment and medium and a target detection model training method, device and equipment and medium. The scheme comprises the following steps: acquiring a to-be-processed question sheet image; the question sheet image comprises a page positioning point; the relative position of the page positioning point and the question sheet in the question sheet image is not changed; detecting coordinate data of each page positioning point in the question sheet image from the question sheet image by using a target detection model; and based on the coordinate data of each page positioning point, performing paging processing on the question sheet image to obtain a question sheet paging result. Therefore, the content in the question sheet can be prevented from being cut off when the question sheet is segmented.

Description

Technical Field

[0001] The embodiments of this specification relate to the field of computer technology, and in particular to a question sheet paging method, device, equipment and medium, and a target detection model training method, device, equipment and medium. Background Art

[0002] In recent years, with the continuous upgrading and development of information technology, the use of mobile terminal devices has become more and more popular, and more convenient, fast and efficient ways of working and learning have become more and more popular. In the traditional field of education, a new generation of educational informatization upgrade exploration has gradually been carried out. In order to reduce the pressure of teachers on marking papers, artificial intelligence (AI) technology is currently used as a new marking method for online marking, which has been widely used.

[0003] In school exam scenarios, A3-style question sheets (test papers or answer sheets) are usually used. After the user obtains the picture of the question sheet by scanning or taking a photo, the picture is input into the AI ​​grading model for processing. However, if the A3-style question sheet is used directly for AI grading, the question sheet picture will be scaled smaller (compared with the A4 style) due to the input size limit of the AI ​​grading model. Therefore, the key information in the A3 question sheet picture is easily lost, which affects the grading accuracy. In order to improve the grading accuracy, the A3-style question sheet in the picture needs to be paginated into two pages of A4-style question sheets or multiple pages, and then the paginated question sheets are processed using the AI ​​grading model.

[0004] In the prior art, the question sheet is segmented at a fixed position on the A3 question sheet image. However, the question sheet may be skewed or slightly truncated during the printing and copying process, resulting in the content in the question sheet not being centered as required. In this case, if fixed-position segmentation is used, important information will be truncated during segmentation, resulting in inaccurate segmentation of the question sheet and affecting the subsequent grading accuracy.

[0005] Based on this, how to avoid truncation of information when splitting and paging the question list has become a technical problem that needs to be solved urgently. Summary of the invention

[0006] In view of this, an embodiment of this specification provides a question list paging method, and an embodiment of this specification also provides a method for training a target detection model. One or more embodiments of this specification simultaneously relate to a question list paging device, a target detection model training device, a computing device, and a computer-readable storage medium to solve the technical defects existing in the prior art.

[0007] According to a first aspect of an embodiment of the present specification, a question sheet paging method is provided, comprising: Acquire a question list image to be processed; the question list image includes a page positioning point; the relative position of the page positioning point and the question list in the question list image remains unchanged; Detecting coordinate data of each of the page positioning points in the question list image from the question list image using a target detection model; Based on the coordinate data of each of the page positioning points, the question list image is paged to obtain a question list paging result.

[0008] According to a second aspect of an embodiment of this specification, a method for training a target detection model is provided, comprising: Acquire a first training sample set; a training sample in the first training sample set includes a first question list image and a first coordinate data set; the first coordinate data set includes coordinate data of each page positioning point whose relative position to the question list in the first question list image remains unchanged; Inputting the first single image into an initial target detection model to obtain a second coordinate data set output by the target detection model; Based on the second coordinate data set and the first coordinate data set, the initial object detection model is trained to obtain a trained object detection model.

[0009] According to a third aspect of the embodiments of this specification, a question sheet paging device is provided, comprising: An acquisition module, used for acquiring a question list image to be processed; the question list image includes a page positioning point; the relative position of the page positioning point and the question list in the question list image remains unchanged; A detection module, used to detect the coordinate data of each of the page positioning points in the question list image from the question list image using a target detection model; The paging module is used to perform paging processing on the question list image based on the coordinate data of each page positioning point to obtain the question list paging result.

[0010] According to a fourth aspect of the embodiments of this specification, there is provided a training device for an icon generation model, comprising: An acquisition module is used to acquire a first training sample set; a training sample in the first training sample set includes a first question list image and a first coordinate data set; the first coordinate data set includes coordinate data of each page positioning point whose relative position to the question list in the first question list image remains unchanged; An input module, used for inputting the first single image into an initial target detection model to obtain a second coordinate data set output by the target detection model; A training module is used to train the initial target detection model based on the second coordinate data set and the first coordinate data set to obtain a trained target detection model.

[0011] According to a fifth aspect of an embodiment of this specification, a computing device is provided, including: Memory and processor; The memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions. When the computer-executable instructions are executed by the processor, the steps of the above-mentioned question list paging method or the steps of the above-mentioned target detection model training method are implemented.

[0012] According to the sixth aspect of the embodiments of this specification, a computer-readable storage medium is provided, which stores computer-executable instructions. When the computer-executable instructions are executed by a processor, the steps of the above-mentioned question list paging method or the steps of the above-mentioned target detection model training method are implemented.

[0013] At least one embodiment provided in the present specification can achieve the following beneficial effects: after obtaining the question list image to be processed including the page positioning points, the target detection model can be used to detect the coordinate data of each page positioning point in the question list image from the question list image, and based on the coordinate data of each page positioning point, the question list image is paged to obtain the question list paging result. Since the relative position of the page positioning point and the question list in the question list image is unchanged, therefore, the question list image is paged based on the coordinate data of each page positioning point. This can avoid truncation of the content in the question list during paging, thereby ensuring the integrity of the question list content and helping to improve the accuracy of question list paging. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] Figure 1 It is a flowchart of a question sheet paging method provided by an embodiment of this specification; Figure 2 is a schematic diagram of a question sheet image provided by an embodiment of this specification; Figure 3 It is a schematic diagram of a question sheet image correction provided by an embodiment of this specification; Figure 4 is a schematic diagram of a preset effective area provided by an embodiment of this specification; Figure 5 is a schematic diagram of a cutting line provided by an embodiment of the present specification; Figure 6 It is a flowchart of a method for training a target detection model provided by an embodiment of this specification; Figure 7 It is a structural schematic diagram of a question sheet paging device provided by an embodiment of this specification; Figure 8 It is a structural schematic diagram of a training device for a target detection model provided by an embodiment of this specification; Fig. 9 It is a structural block diagram of a computing device provided by an embodiment of this specification. DETAILED DESCRIPTION

[0015] Many specific details are described in the following description to facilitate a full understanding of this specification. However, this specification can be implemented in many other ways than those described herein, and those skilled in the art can make similar generalizations without violating the connotation of this specification, so this specification is not limited to the specific implementation disclosed below.

[0016] The terms used in one or more embodiments of this specification are only for the purpose of describing specific embodiments, and are not intended to limit one or more embodiments of this specification. The singular forms of "a", "said" and "the" used in one or more embodiments of this specification and the appended claims are also intended to include plural forms, unless the context clearly indicates other meanings. It should also be understood that the term "and / or" used in one or more embodiments of this specification refers to and includes any or all possible combinations of one or more associated listed items.

[0017] It should be understood that although the terms first, second, etc. may be used to describe various information in one or more embodiments of this specification, this information should not be limited to these terms. These terms are only used to distinguish the same type of information from each other. For example, without departing from the scope of one or more embodiments of this specification, the first may also be referred to as the second, and similarly, the second may also be referred to as the first. Depending on the context, the word "if" as used herein may be interpreted as "at the time of" or "when" or "in response to determining".

[0018] In addition, it should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in one or more embodiments of this specification are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with the relevant laws, regulations and standards of the relevant countries and regions, and provide corresponding operation entrances for users to choose to authorize or refuse.

[0019] In this specification, a question sheet paging method is provided, and an embodiment of this specification also provides a method for training a target detection model. One or more embodiments of this specification simultaneously involve a question sheet paging device, a target detection model training device, a computing device, and a computer-readable storage medium, which are described in detail one by one in the following embodiments.

[0020] See also Figure 1 , Figure 1 This is a flowchart of a question sheet paging method provided by an embodiment of this specification. From a program perspective, the execution subject of the process can be a question sheet paging program. From a hardware perspective, the execution subject of the process can be a device or equipment equipped with a question sheet paging program.

[0021] like Figure 1 As shown, the process may include the following steps: Step 102: Acquire a question list image to be processed; the question list image includes a page positioning point; the relative position of the page positioning point and the question list in the question list image remains unchanged.

[0022] In the embodiments of this specification, the question list may refer to a test paper or an answer sheet, and the question list image may be an image obtained by a user scanning or photographing the test paper or the answer sheet. In practical applications, the page positioning points may be pre-printed when the question list is printed, that is, the scanned or photographed question list includes the page positioning points, so that the question list image obtained by photographing the question list including the page positioning points also includes the page positioning points, and since these page positioning points are printed when the question list is printed, the relative position of the page positioning points and the question list in the question list image remains unchanged.

[0023] In actual applications, users can scan or take photos of completed paper test papers or paper answer sheets to obtain a question list image, and then upload the question list image to the online grading system for online grading. Online grading can use artificial intelligence technology for online grading, or it can still use manual grading, and there is no specific limitation on this. In actual applications, users can also scan or take photos of question lists that have been manually graded and scored to obtain question list images, and then upload the question list images to the online grading system for online grading review.

[0024] In the embodiments of this specification, the page positioning point can be a black rectangular strip, a black circular dot, or other shapes or forms. As long as the center position coordinates can be identified by the target detection model, there is no specific limitation on the shape and size of the page positioning point.

[0025] Figure 2 FIG. 1 is a schematic diagram of a question list image provided by an embodiment of this specification. Figure 2As shown, the question sheet image contains 6 page positioning points, which are in the shape of black rectangular strips, 4 of which are distributed near the four corner points of the question sheet image, and 2 of which are distributed near the center line of the question sheet image. The four-point area surrounded by the 2 page positioning points on the left side of the question sheet image and the 2 page positioning points in the middle part of the question sheet image contains all the question sheet contents in the left column of the question sheet; the four-point area surrounded by the 2 page positioning points on the right side of the question sheet image and the 2 page positioning points in the middle part of the question sheet image contains all the question sheet contents in the right column of the question sheet. Therefore, when paging the question sheet based on the page positioning points, the content of the question sheet can be avoided from being segmented.

[0026] Step 104: Utilizing the target detection model, the coordinate data of each of the page positioning points in the question sheet image is detected from the question sheet image.

[0027] In the embodiments of this specification, the target detection model is an important technology in the field of computer vision, which is used to automatically identify and locate objects of interest (such as people, vehicles, animals, etc.) in images or videos. The target detection model can not only identify the categories of objects present in the image, but also determine the precise location of these objects in the image. The target detection model has a wide range of applications in many fields, such as autonomous driving, security monitoring, medical image analysis, face recognition, etc. With the continuous development of deep learning technology, the accuracy and efficiency of target detection models are also constantly improving. Common target detection models may include: Object detection models based on region proposals: such as R-CNN, Fast R-CNN, Faster R-CNN, etc. These models achieve object detection by generating candidate regions and performing classification and regression; Object detection models based on single detection: such as YOLO (You Only Look Once) and SSD (Single Shot MultiBox Detector), these models regard the object detection task as a single forward pass process, thereby improving the object detection speed.

[0028] In practical applications, the YOLO model can be trained to obtain a target detection model that has the ability to identify the coordinate data of each page positioning point in the question list image. Among them, YOLO is a real-time target detection algorithm. Compared with traditional target detection algorithms, YOLO regards the target detection task as a single regression problem, directly mapping from image pixels to bounding box coordinates and category probabilities. This method greatly simplifies the target detection process and significantly improves the detection speed. The core idea of ​​YOLO is to divide the image into a grid, and each grid unit is responsible for detecting the target whose center point falls within the unit. Each grid unit predicts a fixed number of bounding boxes and the confidence scores of these bounding boxes (indicating whether there is a target in the bounding box and the accuracy of the target). In addition, YOLO also predicts a set of category probabilities for each bounding box, indicating the possibility that the target in the bounding box belongs to each category.

[0029] On the one hand, YOLO regards target detection as a single regression problem, and can obtain the detection result through a single forward propagation, so it has a very high detection speed and can meet the needs of real-time applications. On the other hand, YOLO uses the information of the entire image when predicting, not just the information of the candidate area, which enables YOLO to better utilize contextual information and improve the accuracy of detection. In addition, YOLO can also learn the common features of the target and has good generalization ability when facing new scenes or new targets. It should be noted that other types of target detection models can also be trained to obtain a target detection model that has the ability to identify the coordinate data of each page positioning point in the question list image in the question list image, and there is no specific limitation on this.

[0030] In the embodiments of the present specification, the question list image can be used as the input of the target detection model. The target detection model can detect the question list image of the input model, detect the page positioning points in the question list image, and establish a two-dimensional rectangular coordinate system with a corner point of the question list image (which can be fixed as the upper left corner point or other corner points, without specific limitation) as the coordinate origin, and based on the rectangular coordinate system, determine the coordinate data of each page positioning point in the question list image in the rectangular coordinate system.

[0031] Step 106: Based on the coordinate data of each of the page positioning points, the question list image is paged to obtain a question list paging result.

[0032] In practical applications, the question sheet in the question sheet image may be double-column, triple-column, or more than three-column. For a double-column question sheet, a dividing line can be determined based on the coordinate data of each page positioning point in the question sheet image, and the question sheet image can be paginated to obtain two question sheet images, each of which contains one column of question sheet content; for a three-column question sheet, two dividing lines can be determined based on the coordinate data of each page positioning point in the question sheet image, and the question sheet image can be paginated to obtain three question sheet images, each of which contains one column of question sheet content, and the same applies to the others.

[0033] Figure 1 According to the method in, after obtaining the question list image to be processed including the page positioning points, the target detection model can be used to detect the coordinate data of each page positioning point in the question list image from the question list image, and based on the coordinate data of each page positioning point, the question list image is paged to obtain the question list paging result. Since the relative position of the page positioning point and the question list in the question list image is unchanged, therefore, the question list image is paged based on the coordinate data of each page positioning point. This can avoid truncation of the content in the question list during paging and segmentation, thereby ensuring the integrity of the question list content and helping to improve the accuracy of question list paging.

[0034] based on Figure 1 The method in this specification also provides some specific implementation plans of the method, which are described below.

[0035] As an implementation mode, optionally, in the embodiments of this specification, before the use of the target detection model to detect the coordinate data of each of the page positioning points in the question list image from the question list image, the method may also include: Detecting the question sheet image to obtain coordinate data of four corner points of the question sheet page in the question sheet image; Determine a target transformation matrix based on the coordinate data of the four corner points of the question sheet page and the coordinate data of the four corner points of the question sheet image; Based on the target transformation matrix, the question sheet page is transformed to obtain a corrected question sheet image.

[0036] In actual applications, the user takes a photo of the question list page to obtain a question list image. The question list page may only be located in a part of the question list image, that is, the question list image may include other irrelevant background areas in addition to the question list page. In addition, the question list page may be tilted, appear larger when near and smaller when far away. These situations may have an adverse effect on image recognition and may even cause image recognition failure or inaccurate recognition. Therefore, the question list image needs to be corrected.

[0037] In practical applications, the trapezoidal correction technology can be used to correct the question sheet page. Among them, the trapezoidal correction technology is an image processing technology used to correct the image shape distortion caused by improper shooting angle or position, so that the image is transformed from a trapezoid or other irregular shape to a rectangle to eliminate errors and deformations. This technology has a wide range of applications in computer vision, machine vision, image processing and other fields. The basic principle of trapezoidal correction is to transform the trapezoidal distorted image into a rectangular image by performing a perspective transformation on the image, thereby eliminating the tilt and distortion in the image.

[0038] In the embodiment of this specification, during the correction of the question sheet image, the question sheet image can be detected first, the outline of the question sheet page in the question sheet image can be detected, and the coordinate data of the four corner points of the question sheet page can be determined. Since the subsequent embodiments in the embodiment of this specification will explain in detail the process of determining the coordinate data of the four corner points of the question sheet page, it will not be repeated here.

[0039] In an embodiment of the present specification, after determining the coordinate data of the four corner points of the question list page in the question list image, a target transformation matrix can be determined based on the coordinate data of the four corner points of the question list page and the coordinate data of the four corner points of the question list image. The target transformation matrix is ​​used to adjust the question list page area in the question list image to the same size as the question list image, that is, after converting the question list page based on the target transformation matrix, a corrected question list image can be obtained. The corrected question list image contains only the question list page, thereby removing other irrelevant background areas, and correcting the position and shape of the question list page.

[0040] Figure 3 FIG. 1 is a schematic diagram of a single image correction method provided by an embodiment of the present specification. Figure 3 As shown, in the question list image before correction, the question list page only occupies part of the question list image, and the four corner points of the question list image are point A, point B, point C, and point D, respectively, while the four corner points of the question list page are point a, point b, point c, and point d, respectively. Therefore, the target transformation matrix can be calculated based on the coordinates of the eight points A, point B, point C, point D, point a, point b, point c, and point d, and then the question list page is transformed based on the target transformation matrix to obtain the corrected question list image. The corrected question list image contains the question list page and does not contain other irrelevant background areas. The four corner points of the corrected question list page coincide with the four corner points of the question list image, that is, point a coincides with point A, point b coincides with point B, point c coincides with point C, and point d coincides with point D.

[0041] As an implementation manner, optionally, in the embodiments of this specification, the detecting of the question list image to obtain the coordinate data of four corner points of the question list page in the question list image may specifically include: Converting the question sheet image into a grayscale image; Detecting the edge of the question sheet page in the grayscale image using an edge detection algorithm to obtain the outline of the question sheet page; The contour of the question sheet page is processed using a contour estimation function to obtain coordinate data of four corner points of the question sheet page.

[0042] In the embodiments of this specification, a grayscale image is an image that contains only brightness information but not color information, and the value of each pixel can represent different brightness levels, from black (0) to white (255) or black (0) to white (1), depending on the pixel representation method used. In target detection and recognition, grayscale images can reduce computational complexity and increase processing speed. Therefore, converting the question list image into a grayscale image can increase the processing speed of subsequent detection of the question list image.

[0043] In practical applications, modules in the open source library OpenCV can be used to complete the correction task for the question image. Among them, OpenCV (Open Source Computer Vision Library) is an open source computer vision and machine learning software library, which provides a wealth of functions and tools for a variety of computer vision tasks such as real-time image processing, video analysis, feature recognition, machine learning and deep learning.

[0044] In the embodiments of this specification, an edge detection algorithm is used to detect the edge of the question sheet page in the grayscale image to obtain the outline of the question sheet page. Specifically, the image reading function in OpenCV (such as the cv2.imread function) can be used to read the question sheet image that needs to be corrected and convert the question sheet image into a grayscale image.

[0045] In the embodiments of this specification, after converting the question sheet image into a grayscale image, an edge detection algorithm (such as the Canny algorithm) can be used to detect the edge of the question sheet page in the grayscale image. Since the edge detection algorithm can identify the location where the brightness changes dramatically in the image, the outline of the question sheet page can be determined.

[0046] Specifically, after determining the outline of the question sheet page, you can use the cv2.approxPolyDP function in OpenCV to process the outline of the question sheet page to obtain the coordinate data of the four corner points of the question sheet page. Among them, the cv2.approxPolyDP function is a contour estimation function in OpenCV. The cv2.approxPolyDP function uses the Douglas-Peucker algorithm to simplify the contour and approximate the shape of the original contour by deleting unimportant points. In actual applications, the contour of the question sheet page is input into the cv2.approxPolyDP function. The cv2.approxPolyDP function will approximate the input contour according to the specified accuracy parameters and output the coordinate data of the four corner points of the question sheet page.

[0047] As an implementation mode, optionally, in the embodiment of this specification, before the paging process of the question list image based on the coordinate data of each of the page positioning points is obtained, the following steps may be further included: Based on the paging requirements, a preset number of preset valid areas is determined; According to the detected coordinate data of each of the page anchor points, a page anchor point located in the preset effective area is determined as a valid page anchor point; Correspondingly, the paging process of the question list image is performed based on the coordinate data of each of the page positioning points to obtain the question list paging result, which may specifically include: Based on the coordinate data of each of the valid page positioning points, the question list image is paged to obtain a question list paging result.

[0048] In actual applications, users can set the paging requirements themselves. For example, if the user uploads a three-column question sheet image, the question sheet image can be set to be paginated into 3 pages. Alternatively, the program can determine the corresponding paging requirements based on the result of identifying the number of columns of the question sheet content in the question sheet image. There is no specific limitation on this.

[0049] In the embodiments of this specification, when paging the question list image based on the coordinate data of each page positioning point in the question list image, in order to prevent the situation where the points in the question list content are mistakenly judged as page positioning points, affecting the accuracy of the paging process, a preset number of preset valid areas can be set, and the page positioning points located in the preset valid areas can be determined as valid page positioning points, and the question list image can be paging based on the coordinate data of each valid page positioning point, so that the influence of the page positioning points located in unreasonable areas or the misidentified page positioning points on the accuracy of the paging process of the question list image can be removed. This is beneficial to improving the accuracy of the paging process of the question list image, avoiding truncation of the content in the question list during paging, and is beneficial to ensuring the integrity of the question list content.

[0050] As an implementation manner, optionally, in the embodiment of this specification, the relationship between the preset number of the preset valid areas and the number of pages required is:

[0051] Wherein, M is the preset number of preset valid areas, and N is the number of pages required.

[0052] In the embodiment of the present specification, based on the paging requirement, the preset number of preset valid areas is determined. Specifically, the preset number of preset valid areas can be calculated based on the relationship between the preset number of preset valid areas and the number of required pages according to the number of required pages. For example, if the question sheet image needs to be divided into 3 pages, then N=3 in the formula, and M=8 is calculated by substituting it into the formula, then the preset number of preset valid areas is 8.

[0053] In actual applications, the area position, area shape and area size of each preset valid area can be set and adjusted according to actual needs, and there is no specific limitation on the area position, area shape and area size of the preset valid area. Based on the paging requirements, after the number of preset valid areas is determined, each preset valid area can also be numbered. For example: each preset valid area can be numbered from left to right and from top to bottom, or each preset valid area can be numbered in other ways or orders, and there is no specific limitation on this.

[0054] Figure 4 Schematic diagram of a preset effective area provided by an embodiment of this specification. Figure 4 As shown in FIG. 1 , the question sheet page is double-column and needs to be divided into two pages. Therefore, 6 preset valid areas can be determined accordingly. After numbering each preset valid area from left to right and from top to bottom, the positions of preset valid area 1, preset valid area 2, preset valid area 3, preset valid area 4, preset valid area 5, and preset valid area 6 are obtained as shown in FIG. Figure 4as shown in .

[0055] As an implementation mode, optionally, in the embodiment of this specification, the paging process of the question list image based on the coordinate data of each valid page positioning point is performed to obtain the question list paging result, which may specifically include: According to the paging requirement, determining at least one group of target valid areas for paging in the preset valid areas; Based on the coordinate data of the valid page positioning points existing in each group of target valid areas in the at least one group of target valid areas, determining the segmentation lines corresponding to each group of target valid areas respectively; The question sheet image is paginated using the dividing line to obtain a question sheet pagination result.

[0056] In the embodiments of this specification, the number of dividing lines that need to be determined varies depending on the paging requirements. For example: if the question list image needs to be divided into two pages, only one dividing line needs to be determined; if the question list image needs to be divided into three pages, two dividing lines need to be determined; if the question list image needs to be divided into four pages, three dividing lines need to be determined, and so on. Each dividing line can correspond to a group of target valid areas, and each group of target valid areas is a group of preset valid areas corresponding to a certain dividing line determined from multiple preset valid areas. Therefore, if the question list image needs to be divided into two pages, a group of target valid areas needs to be determined; if the question list image needs to be divided into three pages, two groups of target valid areas need to be determined; if the question list image needs to be divided into four pages, three groups of target valid areas need to be determined, and so on.

[0057] In the embodiments of this specification, for any set of target valid areas, the segmentation lines corresponding to the set of target valid areas can be calculated based on the coordinate data of the valid page positioning points in the set of target valid areas. For a question list image that needs to be paginated using multiple segmentation lines, all segmentation lines can be determined and then the question list image can be paginated using the segmentation lines, or each time a new segmentation line is determined, the question list image can be segmented once using the newly determined segmentation line, and no specific limitation is made to this.

[0058] Figure 5 Schematic diagram of a cutting line provided by an embodiment of this specification. Figure 5As shown, the question sheet page is double-column and needs to be divided into three pages. Therefore, two dividing lines (dividing line 1 and dividing line 2) are needed to perform paging on the question sheet page. Among them, preset valid area 1, preset valid area 2, preset valid area 3, preset valid area 5, preset valid area 6 and preset valid area 7 are a group of target valid areas. The group of target valid areas corresponds to dividing line 1. Through the coordinate data of valid page positioning point 1, valid page positioning point 2, valid page positioning point 3, valid page positioning point 5, valid page positioning point 6 and valid page positioning point 7 in the group of target valid areas, the coordinates of the points on dividing line 1 can be calculated, thereby determining dividing line 1. The preset valid area 2, the preset valid area 3, the preset valid area 4, the preset valid area 6, the preset valid area 7 and the preset valid area 8 are another group of target valid areas, which correspond to the dividing line 2. Through the coordinate data of the valid page positioning point 2, the valid page positioning point 3, the valid page positioning point 4, the valid page positioning point 6, the valid page positioning point 7 and the valid page positioning point 8 in the group of target valid areas, the coordinates of the points on the dividing line 2 can be calculated, thereby determining the dividing line 2.

[0059] As an implementation manner, optionally, in the embodiment of this specification, the determining of the segmentation lines corresponding to each group of target valid areas based on the coordinate data of the valid page positioning points existing in each group of target valid areas among the at least one group of target valid areas may specifically include: For any group of target valid areas among the at least one group of target valid areas, determine whether the valid page positioning point exists in a first valid area of ​​the group of target valid areas, and obtain a first determination result; the first valid area is a preset valid area located in the middle part of the group of target valid areas; If the first judgment result indicates that the valid page anchor point exists in the first valid area, a segmentation line corresponding to any group of target valid areas is determined according to the coordinate data of the valid page anchor point in the first valid area.

[0060] In practical applications, a group of target valid areas can usually include six valid areas, among which the middle part can include two valid areas, the left part can include two valid areas, and the right part can include two valid areas. If there is at least one valid page anchor point in the two valid areas of the middle part, the dividing line corresponding to the group of target valid areas can be directly determined based on the coordinate data of the valid page anchor point in the valid area of ​​the middle part. If there is no valid page anchor point in either of the two valid areas of the middle part, the dividing line corresponding to the group of target valid areas can be determined based on the coordinate data of the valid page anchor point in the valid area of ​​the left part and the valid area of ​​the right part. Therefore, in order to improve the calculation efficiency of the dividing line, before determining the dividing line, it can be judged whether there is a valid page anchor point in the preset valid area of ​​the middle part of each group of target valid areas, and different dividing line determination strategies can be used based on the judgment result.

[0061] In the embodiment of the present specification, if there is a valid page positioning point in the first valid area, when determining the dividing line based on the coordinate data of the valid page positioning point in the first valid area, the average of the horizontal coordinates of each valid page positioning point can be determined as the horizontal coordinate of the dividing line based on the horizontal coordinate data of each valid page positioning point in the first valid area to obtain the dividing line. In this case, the determined dividing line is a line perpendicular to the horizontal axis. Alternatively, when the first valid area includes two preset valid areas, and there are valid page positioning points in each preset valid area, the horizontal coordinate and the vertical coordinate of the first dividing point can also be calculated based on the horizontal coordinates and vertical coordinates of all the valid page positioning points in one of the preset valid areas (specifically, the horizontal coordinates of each valid page positioning point can be averaged to obtain the horizontal coordinate of the first dividing point, and the vertical coordinates of each valid page positioning point can be averaged to obtain the vertical coordinate of the first dividing point). Similarly, the horizontal coordinate and the vertical coordinate of the second dividing point in another preset valid area can be calculated, and the dividing line can be obtained by connecting the first dividing point and the second dividing point.

[0062] As an implementation manner, optionally, in the embodiment of this specification, the determining whether the valid page anchor point exists in the first valid area of ​​any group of target valid areas, after obtaining the first determination result, may further include: If the first judgment result indicates that the valid page anchor point does not exist in the first valid area, it is judged whether at least one valid page anchor point exists in both the second valid area and the third valid area to obtain a second judgment result; the second valid area is a preset valid area located in the left part of any group of target valid areas; the third valid area is a preset valid area located in the right part of any group of target valid areas; If the third judgment result indicates that at least one of the valid page positioning points in both the second valid area and the third valid area is not established, a prompt message indicating that the paging failed is fed back; If the third judgment result indicates that there is at least one valid page positioning point in both the second valid area and the third valid area, then the dividing line corresponding to any group of target valid areas is determined based on the coordinate data of the valid page positioning point in the second valid area and the coordinate data of the valid page positioning point in the third valid area.

[0063] In the embodiments of the present specification, for any group of target valid areas, if there is no valid page positioning point in the first valid area located in the middle part of the group of target valid areas, it is necessary to further determine whether there is at least one valid page positioning point in the second valid area on the left part and the third valid area on the right part of the group of target valid areas. If there is no valid page positioning point in both the second and third valid areas, or there is a valid page positioning point in only one of the second and third valid areas, and no valid page positioning point in the other area, the dividing line cannot be determined, and the question list image cannot be paged. In this case, a prompt message of paging failure can be fed back to the user to prompt the user to re-upload the question list image.

[0064] In the embodiment of the present specification, if there is at least one valid page positioning point in both the second valid area and the third valid area, the average of the horizontal coordinates of all the valid page positioning points in the second valid area and the third valid area can be determined as the horizontal coordinate of the dividing line to obtain the dividing line. In this case, the determined dividing line is a line perpendicular to the horizontal coordinate axis. Alternatively, when there are valid page positioning points in both the two preset valid areas included in the second valid area, and there are valid page positioning points in both the two preset valid areas included in the third valid area, the horizontal coordinate and vertical coordinate of the first dividing point can be calculated based on the horizontal coordinate and vertical coordinate of the valid page positioning point in the preset valid area located on the upper side of the second valid area, and the horizontal coordinate and vertical coordinate of the valid page positioning point in the preset valid area located on the upper side of the third valid area. Then, based on the horizontal coordinate and vertical coordinate of the valid page positioning point in the preset valid area located on the lower side of the second valid area, and the horizontal coordinate and vertical coordinate of the valid page positioning point in the preset valid area located on the lower side of the third valid area, the horizontal coordinate and vertical coordinate of the second dividing point are calculated. The dividing line can be obtained by connecting the first dividing point and the second dividing point.

[0065] For example, Figure 5 Take this as an example to illustrate: Figure 5As shown, preset effective area 1, preset effective area 2, preset effective area 3, preset effective area 5, preset effective area 6 and preset effective area 7 are a group of target effective areas, and the group of target effective areas corresponds to cutting line 1. In the group of target effective areas, preset effective area 2 and preset effective area 6 belong to the first effective area; preset effective area 1 and preset effective area 5 belong to the second effective area; preset effective area 3 and preset effective area 7 belong to the third effective area. Since there are effective page positioning points in preset effective area 2 and preset effective area 6, cutting line 1 can be determined only based on the coordinate data of the effective page positioning points in preset effective area 2 and preset effective area 6.

[0066] See also Figure 6 , Figure 6 This is a flowchart of a method for training a target detection model provided by an embodiment of this specification. From a program perspective, the execution subject of the process can be a program installed on a server or training platform. From a hardware perspective, the execution subject of the process can be a server or training platform that can perform model training.

[0067] like Figure 6 As shown, the process may include the following steps: Step 602: Obtain a first training sample set; a training sample in the first training sample set includes a first question sheet image and a first coordinate data set; the first coordinate data set includes coordinate data of each page positioning point whose relative position to the question sheet in the first question sheet image remains unchanged.

[0068] Step 604: Input the first single image into an initial target detection model to obtain a second coordinate data set output by the target detection model.

[0069] Step 606: Based on the second coordinate data set and the first coordinate data set, the initial target detection model is trained to obtain a trained target detection model.

[0070] In the embodiment of this specification, the question list image, the question list, the page positioning point and the target detection model can be Figure 1 The meanings of the terms mentioned in the embodiments of the method are consistent. Since the meanings of these terms have been explained in detail above, they will not be repeated here.

[0071] In the embodiments of this specification, the initial target detection model may include but is not limited to: R-CNN model, Fast R-CNN model, Faster R-CNN model, YOLO model, SSD model, etc. Since the initial target detection model does not have the ability to detect and identify page positioning points in the question list image, the initial target detection model can be trained using a first training sample set including the first question list image and the first coordinate data set, and the trained target detection model has the ability to detect and identify page positioning points in the question list image.

[0072] In practical applications, the samples can also be expanded by adding noise to the samples or rotating the sample images, wherein the noise added to the samples can be Gaussian random noise, distortion noise, or compression noise, without specific limitation. By training the target detection model with samples with added noise, the ability of the target detection model to resist noise interference can be improved. By training the target detection model with rotated image samples, the target detection model can be adapted to a variety of image rotation situations, which is conducive to enhancing the generalization performance of the target detection model.

[0073] Figure 6 The method in , by using samples containing the question list image and the coordinate data of each page positioning point whose relative position with the question list in the question list image remains unchanged, is used to train the initial target detection model. Thus, a target detection model with the ability to detect and recognize the page positioning points in the question list image can be obtained, which is beneficial to improving the recognition accuracy of the target detection model for each page positioning point in the question list image.

[0074] based on Figure 6 The method in this specification also provides some specific implementation plans of the method, which are described below.

[0075] As an implementation manner, optionally, in the embodiment of this specification, the training of the initial target detection model based on the second coordinate data set and the first coordinate data set to obtain the trained target detection model may specifically include: Determine a loss value between the first coordinate data set and the second coordinate data set; the loss value is used to reflect the size of the difference between the first coordinate data set and the second coordinate data set; Taking the minimum loss value as the adjustment target, the parameters in the initial target detection model are adjusted to obtain a trained target detection model.

[0076] In the embodiments of this specification, a preset loss function can be used to calculate the loss value between the first coordinate data set and the second coordinate data set, and the parameters in the initial target detection model are adjusted with the minimum loss value as the adjustment target to obtain a trained target detection model. In addition, the loss value is used to reflect the difference between the first coordinate data set and the second coordinate data set, that is, the smaller the difference between the second coordinate data set output by the model and the first coordinate data set input to the model, the better the target detection model is.

[0077] Corresponding to the above method embodiment, this specification also provides a question sheet paging device embodiment, Figure 7 FIG. 2 shows a schematic diagram of a structure of a question sheet paging device provided by an embodiment of the present specification. Figure 7 As shown, the device may include: The acquisition module 702 is used to acquire the question list image to be processed; the question list image includes a page positioning point; the relative position of the page positioning point and the question list in the question list image remains unchanged.

[0078] The detection module 704 is used to detect the coordinate data of each page positioning point in the question list image from the question list image by using the target detection model.

[0079] The paging module 706 is used to perform paging processing on the question list image based on the coordinate data of each page positioning point to obtain the question list paging result.

[0080] Optionally, the device may further include: The corner point coordinate determination module is used to detect the question sheet image and obtain the coordinate data of the four corner points of the question sheet page in the question sheet image.

[0081] The transformation matrix determination module is used to determine the target transformation matrix based on the coordinate data of the four corner points of the question sheet page and the coordinate data of the four corner points of the question sheet image.

[0082] The question sheet page conversion module is used to convert the question sheet page based on the target conversion matrix to obtain a corrected question sheet image.

[0083] Optionally, the corner point coordinate determination module may specifically include: The grayscale image conversion submodule is used to convert the question sheet image into a grayscale image.

[0084] The edge detection submodule is used to detect the edge of the question sheet page in the grayscale image using an edge detection algorithm to obtain the outline of the question sheet page.

[0085] The corner point coordinate determination submodule is used to process the outline of the question list page using a contour estimation function to obtain coordinate data of the four corner points of the question list page.

[0086] Optionally, the device may further include: The effective area determination module is used to determine a preset number of preset effective areas based on paging requirements.

[0087] The effective page anchor point determination module is used to determine the page anchor points located in the preset effective area as effective page anchor points according to the detected coordinate data of each of the page anchor points.

[0088] Correspondingly, the paging module 706 may specifically include: The paging submodule is used to perform paging processing on the question list image based on the coordinate data of each valid page positioning point to obtain the question list paging result.

[0089] Optionally, the relationship between the preset number of the preset valid areas and the number of pages required is:

[0090] Wherein, M is the preset number of preset valid areas, and N is the number of pages required.

[0091] Optionally, the paging submodule may specifically include: The target valid area determination unit is used to determine at least one group of target valid areas for paging in the preset valid area according to the paging requirement.

[0092] The cutting line determining unit is used to determine the cutting lines corresponding to each group of target valid areas based on the coordinate data of the valid page positioning points existing in each group of target valid areas in the at least one group of target valid areas.

[0093] The paging processing unit is used to perform paging processing on the question list image using the cutting line to obtain a question list paging result.

[0094] Optionally, the segmentation line determining unit may specifically include: The first judgment subunit is used to judge whether the valid page positioning point exists in the first valid area of ​​any one group of target valid areas among the at least one group of target valid areas, and obtain a first judgment result; the first valid area is a preset valid area located in the middle part of any one group of target valid areas.

[0095] The first dividing line determining subunit is used to determine the dividing line corresponding to any group of target valid areas according to the coordinate data of the valid page anchor point in the first valid area if the first judgment result indicates that the valid page anchor point exists in the first valid area.

[0096] Optionally, the segmentation line determining unit may further include: The second judgment subunit is used to judge whether at least one of the valid page positioning points exists in the second valid area and the third valid area if the first judgment result indicates that the valid page positioning point does not exist in the first valid area, so as to obtain a second judgment result; the second valid area is a preset valid area located in the left part of any group of target valid areas; the third valid area is a preset valid area located in the right part of any group of target valid areas.

[0097] The prompt information feedback subunit is used to feedback prompt information of paging failure if the third judgment result indicates that at least one of the valid page positioning points in the second valid area and the third valid area is not established.

[0098] The second dividing line determining subunit is used to determine the dividing line corresponding to any group of target valid areas based on the coordinate data of the valid page positioning point existing in the second valid area and the coordinate data of the valid page positioning point existing in the third valid area if the third judgment result indicates that there is at least one valid page positioning point in both the second valid area and the third valid area.

[0099] Corresponding to the above method embodiment, this specification also provides an embodiment of a training device for a target detection model, Figure 8 FIG. 1 is a schematic diagram showing a structure of a training device for a target detection model provided by an embodiment of the present specification. Figure 8 As shown, the device may include: Acquisition module 802 is used to acquire a first training sample set; a training sample in the first training sample set includes a first question list image and a first coordinate data set; the first coordinate data set includes coordinate data of each page positioning point whose relative position with the question list in the first question list image remains unchanged.

[0100] The input module 804 is used to input the first single image into the initial target detection model to obtain a second coordinate data set output by the target detection model.

[0101] The training module 806 is used to train the initial target detection model based on the second coordinate data set and the first coordinate data set to obtain a trained target detection model.

[0102] Optionally, the training module 806 may specifically include: The loss value determination submodule is used to determine the loss value of the first coordinate data set and the second coordinate data set; the loss value is used to reflect the size of the difference between the first coordinate data set and the second coordinate data set.

[0103] The parameter adjustment submodule is used to adjust the parameters in the initial target detection model with the minimum loss value as the adjustment target to obtain a trained target detection model.

[0104] The above is a schematic scheme of a training device for a target detection model of this embodiment. It should be noted that the technical scheme of the training device for the target detection model and the technical scheme of the training method for the target detection model mentioned above belong to the same concept, and the details of the technical scheme of the training device for a target detection model that are not described in detail can be found in the description of the technical scheme of the training method for the target detection model mentioned above.

[0105] Fig. 9 The structure block diagram of a computing device provided by an embodiment of the present specification is shown. The components of the computing device 900 include but are not limited to a memory 910 and a processor 920. The processor 920 is connected to the memory 910 via a bus 930, and the database 950 is used to store data.

[0106] The computing device 900 also includes an access device 940 that enables the computing device 900 to communicate via one or more networks 960. Examples of these networks include a public switched telephone network (PSTN), a local area network (LAN), a wide area network (WAN), a personal area network (PAN), or a combination of communication networks such as the Internet. The access device 940 may include one or more of any type of network interface (e.g., a network interface card (NIC)) that is wired or wireless, such as an IEEE 802.11 wireless local area network (WLAN) wireless interface, a world-wide interoperability for microwave access (Wi-MAX) interface, an Ethernet interface, a universal serial bus (USB) interface, a cellular network interface, a Bluetooth interface, and a near field communication (NFC).

[0107] In one embodiment of the present specification, the above components of the computing device 900 and Fig. 9 Other components not shown in the figure may also be connected to each other, for example, via a bus. It should be understood that Fig. 9 The computing device structure block diagram shown is only for the purpose of illustration, and is not intended to limit the scope of this specification. Those skilled in the art can add or replace other components as needed.

[0108] The computing device 900 may be any type of stationary or mobile computing device, including a mobile computer or mobile computing device (e.g., a tablet computer, a personal digital assistant, a laptop computer, a notebook computer, a netbook, etc.), a mobile phone (e.g., a smart phone), a wearable computing device (e.g., a smart watch, smart glasses, etc.), or other types of mobile devices, or a stationary computing device such as a desktop computer or a personal computer (PC). The computing device 900 may also be a mobile or stationary server.

[0109] Among them, the processor 920 is used to execute the following computer-executable instructions, which, when executed by the processor, implement the steps of the above-mentioned question list paging method or target detection model training method.

[0110] The above is a schematic scheme of a computing device of this embodiment. It should be noted that the technical scheme of the computing device and the technical scheme of the above-mentioned question list paging method or target detection model training method belong to the same concept, and the details not described in detail in the technical scheme of the computing device can be referred to the description of the technical scheme of the above-mentioned question list paging method or target detection model training method.

[0111] An embodiment of the present specification also provides a computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, implement the steps of the above-mentioned question list paging method, or implement the steps of the above-mentioned target detection model training method.

[0112] The above is a schematic scheme of a computer-readable storage medium of this embodiment. It should be noted that the technical scheme of the storage medium and the technical scheme of the above-mentioned question list paging method or target detection model training method belong to the same concept, and the details not described in detail in the technical scheme of the storage medium can be referred to the description of the technical scheme of the above-mentioned question list paging method or target detection model training method.

[0113] The above is a description of a specific embodiment of the specification. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recorded in the claims can be performed in an order different from that in the embodiments and still achieve the desired results. In addition, the processes depicted in the drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0114] The computer instructions include computer program codes, which may be in source code form, object code form, executable files or some intermediate forms, etc. The computer-readable medium may include: any entity or device capable of carrying the computer program code, recording medium, USB flash drive, mobile hard disk, magnetic disk, optical disk, computer memory, read-only memory (ROM), random access memory (RAM), electric carrier signal, telecommunication signal and software distribution medium, etc. It should be noted that the content contained in the computer-readable medium may be appropriately increased or decreased according to the requirements of patent practice. For example, in some regions, according to patent practice, computer-readable media do not include electric carrier signals and telecommunication signals.

[0115] It should be noted that, for the above-mentioned method embodiments, for the sake of simplicity of description, they are all expressed as a series of action combinations, but those skilled in the art should be aware that the embodiments of this specification are not limited by the order of the actions described, because according to the embodiments of this specification, some steps can be performed in other orders or simultaneously. Secondly, those skilled in the art should also be aware that the embodiments described in the specification are all preferred embodiments, and the actions and modules involved are not necessarily required by the embodiments of this specification.

[0116] In the above embodiments, the description of each embodiment has its own emphasis. For parts that are not described in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0117] The preferred embodiments of this specification disclosed above are only used to help explain this specification. The optional embodiments do not describe all the details in detail, nor do they limit the invention to only the specific implementation methods described. Obviously, many modifications and changes can be made according to the content of the embodiments of this specification. This specification selects and specifically describes these embodiments in order to better explain the principles and practical applications of the embodiments of this specification, so that technicians in the relevant technical field can well understand and use this specification. This specification is only limited by the claims and their full scope and equivalents.

Claims

1. A question sheet paging method, characterized in that: The method comprises: Acquire a question list image to be processed; the question list image includes a page positioning point; the relative position of the page positioning point and the question list in the question list image remains unchanged; Detecting coordinate data of each of the page positioning points in the question list image from the question list image using a target detection model; Based on the coordinate data of each of the page positioning points, the question list image is paged to obtain a question list paging result.

2. The method according to claim 1, characterized in that Before detecting the coordinate data of each of the page positioning points in the question list image by using the target detection model, the method further comprises: Detecting the question sheet image to obtain coordinate data of four corner points of the question sheet page in the question sheet image; Determine a target transformation matrix based on the coordinate data of the four corner points of the question sheet page and the coordinate data of the four corner points of the question sheet image; Based on the target transformation matrix, the question sheet page is transformed to obtain a corrected question sheet image.

3. The method according to claim 2, characterized in that The detecting of the question list image to obtain coordinate data of four corner points of the question list page in the question list image specifically includes: Converting the question sheet image into a grayscale image; Detecting the edge of the question sheet page in the grayscale image using an edge detection algorithm to obtain the outline of the question sheet page; The contour of the question sheet page is processed using a contour estimation function to obtain coordinate data of four corner points of the question sheet page.

4. The method according to claim 1, characterized in that Before the paging process of the question list image is performed based on the coordinate data of each of the page positioning points to obtain the question list paging result, the method further includes: Based on the paging requirements, a preset number of preset valid areas is determined; According to the detected coordinate data of each of the page anchor points, a page anchor point located in the preset effective area is determined as a valid page anchor point; The paging process of the question sheet image based on the coordinate data of each of the page positioning points to obtain the question sheet paging result specifically includes: Based on the coordinate data of each of the valid page positioning points, the question list image is paged to obtain a question list paging result.

5. The method according to claim 4, characterized in that The relationship between the preset number of the preset valid areas and the number of pages required is: Wherein, M is the preset number of preset valid areas, and N is the number of pages required.

6. The method according to claim 4, characterized in that The paging process of the question sheet image is performed based on the coordinate data of each valid page positioning point to obtain the question sheet paging result, specifically including: According to the paging requirement, determining at least one group of target valid areas for paging in the preset valid areas; Based on the coordinate data of the valid page positioning point existing in each group of target valid areas in the at least one group of target valid areas, determining the segmentation lines corresponding to each group of target valid areas respectively; The question sheet image is paginated using the dividing line to obtain a question sheet pagination result.

7. The method according to claim 6, characterized in that The determining, based on the coordinate data of the valid page positioning points existing in each group of target valid areas in the at least one group of target valid areas, the segmentation lines corresponding to each group of target valid areas respectively include: For any group of target valid areas among the at least one group of target valid areas, determine whether the valid page positioning point exists in a first valid area of ​​the group of target valid areas, and obtain a first determination result; the first valid area is a preset valid area located in the middle part of the group of target valid areas; If the first judgment result indicates that the valid page anchor point exists in the first valid area, a segmentation line corresponding to any group of target valid areas is determined according to the coordinate data of the valid page anchor point in the first valid area.

8. The method according to claim 7, characterized in that After determining whether the valid page positioning point exists in the first valid area of ​​any group of target valid areas and obtaining the first determination result, the method further includes: If the first judgment result indicates that the valid page anchor point does not exist in the first valid area, it is judged whether at least one valid page anchor point exists in both the second valid area and the third valid area to obtain a second judgment result; the second valid area is a preset valid area located in the left part of any group of target valid areas; the third valid area is a preset valid area located in the right part of any group of target valid areas; If the third judgment result indicates that at least one of the valid page positioning points in both the second valid area and the third valid area is not established, a prompt message indicating that the paging failed is fed back; If the third judgment result indicates that there is at least one valid page positioning point in both the second valid area and the third valid area, then the dividing line corresponding to any group of target valid areas is determined based on the coordinate data of the valid page positioning point in the second valid area and the coordinate data of the valid page positioning point in the third valid area.

9. A method for training a target detection model, characterized in that: The method comprises: Acquire a first training sample set; a training sample in the first training sample set includes a first question list image and a first coordinate data set; the first coordinate data set includes coordinate data of each page positioning point whose relative position to the question list in the first question list image remains unchanged; Inputting the first single image into an initial target detection model to obtain a second coordinate data set output by the target detection model; Based on the second coordinate data set and the first coordinate data set, the initial object detection model is trained to obtain a trained object detection model.

10. The method according to claim 9, characterized in that The training of the initial target detection model based on the second coordinate data set and the first coordinate data set to obtain a trained target detection model specifically includes: Determine a loss value between the first coordinate data set and the second coordinate data set; the loss value is used to reflect the size of the difference between the first coordinate data set and the second coordinate data set; Taking the minimum loss value as the adjustment target, the parameters in the initial target detection model are adjusted to obtain a trained target detection model.

11. A question sheet paging device, characterized in that: The device comprises: An acquisition module, used for acquiring a question list image to be processed; the question list image includes a page positioning point; the relative position of the page positioning point and the question list in the question list image remains unchanged; A detection module, used to detect the coordinate data of each of the page positioning points in the question list image from the question list image using a target detection model; The paging module is used to perform paging processing on the question list image based on the coordinate data of each page positioning point to obtain the question list paging result.

12. A training device for a target detection model, characterized in that: The device comprises: An acquisition module is used to acquire a first training sample set; a training sample in the first training sample set includes a first question list image and a first coordinate data set; the first coordinate data set includes coordinate data of each page positioning point whose relative position to the question list in the first question list image remains unchanged; An input module, used for inputting the first single image into an initial target detection model to obtain a second coordinate data set output by the target detection model; A training module is used to train the initial target detection model based on the second coordinate data set and the first coordinate data set to obtain a trained target detection model.

13. A computing device, characterized in that: include: Memory and processor; The memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions. When the computer-executable instructions are executed by the processor, the steps of the question list paging method described in any one of claims 1 to 8 are implemented, or the steps of the target detection model training method described in any one of claims 9 to 10 are implemented.

14. A computer-readable storage medium, characterized in that: It stores computer executable instructions, which, when executed by a processor, implement the steps of the question list paging method described in any one of claims 1 to 8, or implement the steps of the target detection model training method described in any one of claims 9 to 10.

Images