CN112347998A - Question judging method, device, equipment and storage medium - Google Patents
Question judging method, device, equipment and storage medium Download PDFInfo
- Publication number
- CN112347998A CN112347998A CN202110005089.0A CN202110005089A CN112347998A CN 112347998 A CN112347998 A CN 112347998A CN 202110005089 A CN202110005089 A CN 202110005089A CN 112347998 A CN112347998 A CN 112347998A
- Authority
- CN
- China
- Prior art keywords
- question
- frame
- stem frame
- image
- corrected
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V30/00—Character recognition; Recognising digital ink; Document-oriented image-based pattern recognition
- G06V30/40—Document-oriented image-based pattern recognition
- G06V30/41—Analysis of document content
- G06V30/412—Layout analysis of documents structured with printed lines or input boxes, e.g. business forms or tables
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2415—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on parametric or probabilistic models, e.g. based on likelihood ratio or false acceptance rate versus a false rejection rate
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/20—Image preprocessing
- G06V10/22—Image preprocessing by selection of a specific region containing or referencing a pattern; Locating or processing of specific regions to guide the detection or recognition
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Artificial Intelligence (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Evolutionary Computation (AREA)
- Evolutionary Biology (AREA)
- General Engineering & Computer Science (AREA)
- Bioinformatics & Computational Biology (AREA)
- Life Sciences & Earth Sciences (AREA)
- Multimedia (AREA)
- Probability & Statistics with Applications (AREA)
- Image Analysis (AREA)
Abstract
The application provides a question judging method and device, electronic equipment and a storage medium. The specific implementation scheme is as follows: performing text line detection on the image to be corrected to obtain a question stem frame of a question in the image to be corrected; inputting the text information in the question stem frame into a question number processing model, processing the question number in the question stem frame by using the question number processing model, and outputting an adjusting mode and an adjusting proportion of the question stem frame; adjusting the question stem frame according to the adjusting mode and the adjusting proportion to obtain an adjusted question stem frame; and obtaining a question judging result aiming at the question in the image to be corrected based on the adjusted question stem frame. The embodiment of the application can effectively eliminate the interference of the question number caused by the inaccuracy of the detection frame on the question judgment result, and effectively improve the accuracy of the question judgment result.
Description
Technical Field
The present application relates to the field of artificial intelligence technologies, and in particular, to a method, an apparatus, a device, and a storage medium for determining a question.
Background
The shooting judgment problem is an important application of the artificial intelligence technology in the field of education. The process of judging the subject by taking a picture in a general case can comprise the following steps: and (4) photographing the image to be corrected by using terminal equipment such as a mobile phone or a tablet personal computer and the like, and uploading the image to be corrected to an application program for photographing and judging the question. The application program carries out the processing of multiple links such as text detection, content identification and the like on the image to be corrected, and then outputs the result of judging the question.
The problem of problem number interference exists in the conventional shooting problem judgment method. The topic number interference includes the following cases: due to the inaccuracy of the detection frames of the text detection, the question mark frame may be included in some detection frame results, and the question mark frame may not be included in other detection frame results. The question number interference can greatly interfere with the question judgment accuracy of the question judgment strategy. Because the problem number interference cannot be effectively eliminated, the problem judgment accuracy is influenced to a certain extent, and the use experience of a user is greatly influenced.
Disclosure of Invention
The embodiment of the application provides a method and a device for judging a topic, electronic equipment and a storage medium, which are used for solving the problems in the related art, and the technical scheme is as follows:
in a first aspect, an embodiment of the present application provides a topic determination method, including:
performing text line detection on the image to be corrected to obtain a question stem frame of a question in the image to be corrected;
inputting the text information in the question stem frame into a question number processing model, processing the question number in the question stem frame by using the question number processing model, and outputting an adjusting mode and an adjusting proportion of the question stem frame;
adjusting the question stem frame according to the adjusting mode and the adjusting proportion to obtain an adjusted question stem frame;
and obtaining a question judging result aiming at the question in the image to be corrected based on the adjusted question stem frame.
In one embodiment, the method for processing the question mark in the question stem frame by using the question mark processing model and outputting the adjustment mode and the adjustment proportion of the question stem frame comprises the following steps:
outputting an adjusting mode of the question stem frame by utilizing a first branch of the question mark processing model; the adjusting mode comprises rightward adjustment of the left frame, leftward adjustment or no adjustment of the right frame.
In one embodiment, the method for processing the question mark in the question stem frame by using the question mark processing model and outputting the adjustment mode and the adjustment proportion of the question stem frame comprises the following steps:
and outputting the adjustment proportion of the question stem frame by using the second branch of the question mark processing model.
In one embodiment, the method for detecting a text line of an image to be corrected to obtain a stem frame of a subject in the image to be corrected further includes:
and performing text line detection on the image to be corrected by using the target detection model to obtain a question stem frame and an answer frame of the question in the image to be corrected.
In one embodiment, the method further comprises:
identifying the content of the text information in the question stem frame and the answer frame of the question;
obtaining at least one preliminary question judging result aiming at the question by utilizing at least one question judging model according to the result of the content identification;
and under the condition that all the preliminary question judging results of the questions indicate that the answers of the questions are wrong, executing the step of inputting the text information in the question stem frame into the question number processing model.
In one embodiment, the method further comprises:
and training a second branch of the question mark processing model by adopting an intersection-to-parallel ratio loss function.
In one embodiment, the method further comprises:
adjusting the height coordinate value of the stem frame obtained by the target detection model into the height coordinate value of the real stem frame;
and calculating the intersection ratio of the question stem frame obtained by the target detection model and the real question stem frame.
In a second aspect, an embodiment of the present application provides a topic determination apparatus, including:
the detection unit is used for detecting text lines of the image to be corrected to obtain a question stem frame of a question in the image to be corrected;
the processing unit is used for inputting the text information in the question stem frame into the question mark processing model, processing the question mark in the question stem frame by using the question mark processing model and outputting the adjusting mode and the adjusting proportion of the question stem frame;
the adjusting unit is used for adjusting the question stem frame according to the adjusting mode and the adjusting proportion to obtain an adjusted question stem frame;
and the question judging unit is used for obtaining a question judging result aiming at the question in the image to be corrected based on the adjusted question stem frame.
In one embodiment, the processing unit is configured to:
outputting an adjusting mode of the question stem frame by utilizing a first branch of the question mark processing model; the adjusting mode comprises rightward adjustment of the left frame, leftward adjustment or no adjustment of the right frame.
In one embodiment, the processing unit is configured to:
and outputting the adjustment proportion of the question stem frame by using the second branch of the question mark processing model.
In one embodiment, the detection unit is further configured to:
and performing text line detection on the image to be corrected by using the target detection model to obtain a question stem frame and an answer frame of the question in the image to be corrected.
In one embodiment, the apparatus further includes a preliminary topic determination unit;
the preliminary decision unit is used for: identifying the content of the text information in the question stem frame and the answer frame of the question; obtaining at least one preliminary question judging result aiming at the question by utilizing at least one question judging model according to the result of the content identification;
the processing unit is further configured to: and under the condition that all the preliminary question judging results of the questions indicate that the answers of the questions are wrong, executing the step of inputting the text information in the question stem frame into the question number processing model.
In one embodiment, the apparatus further comprises a training unit; the training unit is used for:
and training a second branch of the question mark processing model by adopting an intersection-to-parallel ratio loss function.
In one embodiment, the training unit is further configured to:
adjusting the height coordinate value of the stem frame obtained by the target detection model into the height coordinate value of the real stem frame;
and calculating the intersection ratio of the question stem frame obtained by the target detection model and the real question stem frame.
In a third aspect, an embodiment of the present application provides an electronic device, including: a memory and a processor. Wherein the memory and the processor are in communication with each other via an internal connection path, the memory is configured to store instructions, the processor is configured to execute the instructions stored by the memory, and the processor is configured to perform the method of any of the above aspects when the processor executes the instructions stored by the memory.
In a fourth aspect, embodiments of the present application provide a computer-readable storage medium, which stores a computer program, and when the computer program runs on a computer, the method in any one of the above-mentioned aspects is executed.
The advantages or beneficial effects in the above technical solution at least include: the method can effectively eliminate the interference of the question numbers caused by inaccurate detection frames to the question judgment results, and effectively improve the accuracy of the question judgment results.
The foregoing summary is provided for the purpose of description only and is not intended to be limiting in any way. In addition to the illustrative aspects, embodiments, and features described above, further aspects, embodiments, and features of the present application will be readily apparent by reference to the drawings and following detailed description.
Drawings
In the drawings, like reference numerals refer to the same or similar parts or elements throughout the several views unless otherwise specified. The figures are not necessarily to scale. It is appreciated that these drawings depict only some embodiments in accordance with the disclosure and are therefore not to be considered limiting of its scope.
FIG. 1 is a flow chart of a topic determination method according to an embodiment of the application;
FIG. 2 is a diagram illustrating a text line detection result of a topic determination method according to another embodiment of the present application;
FIG. 3 is a flow chart of a topic determination method according to another embodiment of the present application;
fig. 4 is a schematic diagram illustrating a stem answer detection result of a question judging method according to another embodiment of the present application;
FIG. 5 is a diagram illustrating a text line detection result of a topic determination method according to another embodiment of the present application;
FIG. 6 is a schematic diagram of a topic determination method according to another embodiment of the present application;
FIG. 7 is a diagram illustrating layout analysis results of a topic determination method according to another embodiment of the present application;
FIG. 8 is a flow chart of model training for a method of determining a subject according to another embodiment of the present application;
FIG. 9 is a flowchart of a topic determination method according to another embodiment of the present application;
FIG. 10 is a schematic structural diagram of a topic determination device according to an embodiment of the present application;
FIG. 11 is a schematic structural diagram of a topic determination device according to another embodiment of the present application;
FIG. 12 is a schematic structural diagram of a topic determination device according to another embodiment of the present application;
FIG. 13 is a block diagram of an electronic device used to implement embodiments of the present application.
Detailed Description
In the following, only certain exemplary embodiments are briefly described. As those skilled in the art will recognize, the described embodiments may be modified in various different ways, all without departing from the spirit or scope of the present application. Accordingly, the drawings and description are to be regarded as illustrative in nature, and not as restrictive.
Fig. 1 is a flowchart of a topic determination method according to an embodiment of the present application. As shown in fig. 1, the method for determining a topic may include:
step S110, performing text line detection on the image to be corrected to obtain a question stem frame of a question in the image to be corrected;
step S120, inputting the text information in the question stem frame into a question number processing model, processing the question number in the question stem frame by using the question number processing model, and outputting the adjusting mode and the adjusting proportion of the question stem frame;
step S130, adjusting the question stem frame according to the adjusting mode and the adjusting proportion to obtain an adjusted question stem frame;
and step S140, obtaining a question judging result aiming at the question in the image to be corrected based on the adjusted question stem frame.
In the related method for judging the photo problem, the problem of problem number interference may exist. Common inscriptions are "(1)", "(11)", "1.", "" 11. "," (r) "and" ("c"), etc. Due to the inaccuracy of the detection frames of the text detection, the question mark frame may be included in some detection frame results, and the question mark frame may not be included in other detection frame results. Fig. 2 is a schematic diagram of a text line detection result of a topic determination method according to another embodiment of the present application. As shown in fig. 2, the test frame result of the subject 1 has the question number frame therein, and the test frame result of the subject 5 has no question number frame therein.
The question number interference can greatly interfere with the question judgment accuracy of the question judgment strategy. For example, for the topic of 1.22+12=13.22, if "1" is the topic number, the topic should be misjudged; if "1.22" is a decimal, the question should be judged. In the problem judgment strategy, it is assumed that all the problems are defaulted to have problem numbers, then the problem numbers are removed for judgment, or all the problems are treated as no problem numbers, and the two strategies can not ensure that the problem of problem number interference can be solved well.
In view of this, an embodiment of the present application provides a method for determining a question, which adjusts a question stem frame of a question in an image to be corrected to eliminate question number interference, and determines the question based on the adjusted question stem frame, thereby effectively improving the accuracy of a question determination result.
In step S110, the image to be corrected may be input to a pre-trained machine learning model, and text line detection is performed on the image to be corrected by using the machine learning model, so as to obtain text region coordinates of a text box included in the image to be corrected. Wherein, one part of the text box is a question stem box of the question, and the other part of the text box is an answer box of the question. In one example, after text line detection is performed on the image to be modified, the image to be modified can be detected by using the stem answer detection model or the layout analysis model. The question stem answer detection model and the layout analysis model may also be pre-trained machine learning models, such as neural network models. And combining the results of the two detections to obtain the text regions of the question stem frame and the answer frame corresponding to each question in the image to be corrected.
In step S120, the text content in the stem frame in the image to be corrected obtained in step S110 may be identified first to obtain the text information in the stem frame. And then inputting the text information in the question stem box into a pre-trained question mark processing model.
In one example, there may be some question stem frames obtained in step S110 that have question numbers in them, and some that do not have question numbers in them. The adjustment mode and the adjustment proportion of the question stem frame can be output by using the question mark processing model, so that all the adjusted question stem frames do not have question mark frames in the question stem frames. Referring to the example of fig. 2, the adjustment mode for the question number 1 output by the question number processing model is that the left frame is adjusted to the right, and the question number is removed from the question stem frame; the adjustment mode of the question number processing model output to the 5 th question is not adjusted. The problem of question number interference is eliminated through the adjustment, and the question is judged based on the adjusted question stem frame in the subsequent steps to obtain an accurate question judgment result.
In another example, the question mark processing model may also be used to output the adjustment mode and the adjustment scale of the question stem frames, so that all the adjusted question stem frames have the question mark therein. In the subsequent question judging strategy, all the questions can be defaulted to have question numbers, and then the question numbers are removed for judging the questions. The problem of question number interference is eliminated through the adjustment, and the question is judged based on the adjusted question stem frame, so that an accurate question judgment result can be obtained.
In step S130, the question stem frames may be adjusted according to the adjustment method and the adjustment ratio obtained in step S120, and the obtained adjusted question stem frames may have no question mark frame in all question stem frames or have question mark frames in all question stem frames.
In step S140, if no question number frame is included in all the adjusted question stem frames, all the questions are treated as no question number in the subsequent question judgment strategy. If all the adjusted question stem frames have the question number frames in them, then in the subsequent question judging strategy, the question numbers are removed and then the question is judged. And obtaining an accurate question judgment result based on the adjusted question stem frame.
The embodiment of the application can effectively eliminate the interference of the question number caused by the inaccuracy of the detection frame on the question judgment result, and effectively improve the accuracy of the question judgment result.
In one embodiment, the method for processing the question mark in the question stem frame by using the question mark processing model and outputting the adjustment mode and the adjustment proportion of the question stem frame comprises the following steps:
outputting an adjusting mode of the question stem frame by utilizing a first branch of the question mark processing model; the adjusting mode comprises rightward adjustment of the left frame, leftward adjustment or no adjustment of the right frame.
Referring to the example of fig. 2, the adjustment manner of the stem frame may be output by using the question mark processing model, so that all the adjusted stem frames do not have the question mark frame therein. The adjustment mode for the 1 st question output by the question number processing model is that the left frame is adjusted rightwards, and the question number is removed from the question stem frame; the adjustment mode of the question number processing model output to the 5 th question is not adjusted.
In another example, if a question stem frame has the question number frame of the next question to the right of the corresponding question in it, for example, the question stem frame of the question 1 in fig. 2 mistakenly has the question number frame of the question 5 in it, the adjustment mode for the question 1 output by the question number processing model is to adjust the right frame to the left and remove the question number from the question stem frame.
In another embodiment, the question mark processing model may also be used to output the adjustment mode and the adjustment ratio of the question stem frames, so that all the adjusted question stem frames have the question mark therein. In this manner, the adjustment of the output of the question mark processing model may include left frame adjustment, right frame adjustment, or no adjustment.
On the other hand, in the topic determination strategy, if all topics are treated as no topic numbers, as in the case of topic 1 in fig. 2, the detection frame has multiple frames of topic numbers, which may cause topic number interference. On the other hand, in the question judgment strategy, if all questions are regarded as having question numbers, and the question numbers are removed to perform question judgment processing, question number interference is also caused due to missing question numbers in the detection frame as in the case of the 5 th question in fig. 2. The problem of problem number interference caused by detecting multiple frames or missing frames of the frame can be effectively solved, and therefore the accuracy of the problem judgment result is effectively improved.
In one embodiment, the method for processing the question mark in the question stem frame by using the question mark processing model and outputting the adjustment mode and the adjustment proportion of the question stem frame comprises the following steps:
and outputting the adjustment proportion of the question stem frame by using the second branch of the question mark processing model.
In order to eliminate the problem of question mark interference, the height of the question stem frame does not need to be adjusted, and only the width of the question stem frame needs to be adjusted. That is, the upper frame and the lower frame of the stem frame do not need to be adjusted, and only the left frame and/or the right frame of the stem frame need to be adjusted. The adjustment proportion of the stem frame can comprise the ratio of the width adjustment size of the stem frame to the width of the stem frame. For example, the width of the question stem frame is 5cm, in which the question number of the question is enclosed on the left side in the question stem frame, and the width occupied by the question number is 1cm, and the adjustment ratio of the question stem frame is 20% by removing the question number from the question stem frame by adjustment. For another example, the width of the question stem frame is 200 pixels, in which the question mark of the question is enclosed on the left side in the question stem frame, and the width occupied by the question mark is 30 pixels, and the adjustment ratio of the question stem frame is 15% by removing the question mark from the question stem frame by the adjustment.
In the embodiment of the present application, the question mark processing model may be composed of two branches. Each branch may consist of 1 pre-convolutional layer and 4 blocks, respectively. Wherein each block may be composed of 3 convolutional layers. The pre-convolution layer of each branch may employ a 3 x 3 convolution kernel.
In one example, the convolution layer in each block of the first branch of the question mark processing model extracts image features using a 1 x 3 convolution kernel. Using a 1 x 3 convolution kernel may focus on extracting the lateral features of the image. The convolution layer in each block of the second branch of the question mark processing model extracts image features using a 3 x 1 convolution kernel. The 3 x 1 convolution kernel may focus on extracting longitudinal features. The number of pre-convolution channels per branch is 32. The first block of each branch has a channel number of 64, the second block has a channel number of 128, the third block has a channel number of 256, and the fourth block has a channel number of 64. And in two branches of the question mark processing model, the output of each block is spliced into an image with higher resolution in an interpolation mode and used as the input of the next block. Wherein the interpolation includes using gray values of known neighboring pixels (or tristimulus values in an RGB image) to generate gray values of unknown pixels so as to reproduce an image with higher resolution from the original image. The image is processed through the network structure, and the global features with stronger expression capability are extracted, so that the obtained image information is more comprehensive.
And finally, in the network output part of the two branches, each branch is respectively connected with 2 full connection layers. The number of nodes at the first fully connected level of each branch is taken to be 256, for example. The second number of fully connected nodes of the first branch is for example taken to be 3. Three nodes respectively represent the adjustment modes of 3 question stem frames: the first indicates that the left frame is adjusted to the right, the second indicates that the right frame is adjusted to the left, and the third indicates that no adjustment is made. The number of the second fully connected nodes of the second branch is 1, and the number represents the adjustment proportion of the question stem frame.
Fig. 3 is a flowchart of a topic determination method according to another embodiment of the present application. As shown in fig. 3, in an embodiment, in step S110 in fig. 1, performing text line detection on the image to be corrected to obtain a stem frame of a topic in the image to be corrected, which may further include:
and step S210, performing text line detection on the image to be corrected by using the target detection model to obtain a stem frame and an answer frame of the question in the image to be corrected.
In one example, the centrnet model may be utilized to perform text line detection on the image to be corrected, resulting in text region coordinates of a text box included in the image to be corrected. CenterNet, also known as Objects as Points, employs a general object detection method. In the object detection method, the class N of the population of objects to be predicted is set first, and finally the number of output channels is N +2+ 2. Including predicting the center point of the object, and outputting one score map for each category, so that there are N score maps. The predicted value corresponding to each pixel point is between 0 and 1, and the probability that the point is the center of a certain type of object is represented. In the prediction process, it cannot be guaranteed that the predicted central point is the real central point, and the predicted value is often deviated in the actual situation, so that the deviation of the central point is predicted by using two channels. One of the offsets is an x-axis offset and the other is a y-axis offset. In addition, the remaining two channels are used to predict the distance of the center point from the left and upper borders of the rectangular box. After the model is used for prediction, a possible central point of an object is found in the score map by setting a threshold, the central point is corrected according to the x-axis offset and the y-axis offset corresponding to the central point, and then the final target detection result is obtained through the central point and the predicted width and height of the rectangular frame.
In step S210, the image to be corrected may be input to a pre-trained centret model, and text line detection is performed on the image to be corrected by using the centret model, so as to obtain text region coordinates of a text box included in the image to be corrected. Wherein, one part of the text box is a question stem box of the question, and the other part of the text box is an answer box of the question.
In one example, after text line detection is performed on the image to be corrected, the stem answer detection model can be used for detecting text region coordinates of a stem frame and text region coordinates of an answer frame in the image to be corrected. And combining the detection result of the question stem answer detection model and the detection result of the text line detection to obtain a question stem frame and a text area of an answer frame corresponding to each question in the image to be corrected.
Specifically, the question stem answer detection result and the text line detection result may be combined by combining means and a position relationship, so as to obtain a question stem frame and a text region of an answer frame corresponding to each question in the image to be corrected. The position relation comprises the position relation between the question stem and the corresponding answer.
Fig. 4 is a diagram illustrating a question stem answer detection result of a question judging method according to another embodiment of the present application. The dark rectangle text box in fig. 4 represents the text area of the stem box in the image to be corrected, which is identified by the stem answer detection model. The light-colored rectangular text box in fig. 4 represents the text area of the answer box in the image to be corrected, which is identified by the question stem answer detection model. Fig. 5 is a schematic diagram of a text line detection result of a topic determination method according to another embodiment of the present application. Referring to fig. 4 and 5, the intersection ratio may be used to establish a corresponding relationship between the answer detection result of the stem and the text line detection result. If the intersection ratio of the text box in the answer detection result of the question stem and the text box in the text line detection result is greater than or equal to a preset second threshold value, the two text boxes can be considered to belong to the same question, and a corresponding relationship can be established between the two text boxes. And then obtaining a question stem frame and a text area of an answer frame corresponding to each question in the image to be corrected according to the position relation between the question stem and the corresponding answer.
The position relationship between the question stem and the corresponding answer is usually related to the question type. For example, the themes may include horizontal, vertical, and diagonal. Fig. 6 is a schematic diagram of a topic determination method according to another embodiment of the present application. In fig. 6, the question corresponding to the question with the question number 1) is horizontal, the question corresponding to the question with the question number 2) is vertical, and the question corresponding to the question with the question number 3) is off-line. For the horizontal-type theme, the positional relationship may include the following situations. First, the text box in the text line detection result may include all of at least one of the stem box and the answer box in the stem answer detection result. Second, a text box in the text line detection result may intersect with a majority of an area of at least one of a stem box and an answer box in the stem answer detection result. Third, the positions of the text box in the text line detection result and the stem box and the answer box in the stem answer detection result may be on the same horizontal line. For upright and drop-off types, the positional relationship may include the position of the answer frame below the stem frame. And intuitively, the question stem box and the answer box are obviously two text boxes.
And obtaining a question stem frame and a text area of an answer frame corresponding to each question in the image to be corrected based on the position relation and the corresponding relation between the question stem answer detection result and the text line detection result.
In another example, after text line detection of the image to be corrected, the image to be corrected may be detected by using the layout analysis model. And combining the layout analysis detection result and the text line detection result by utilizing the cross-over ratio to obtain the text areas of the question stem frame and the answer frame corresponding to each question in the image to be corrected.
The layout analysis model is a text detection model. The layout analysis detection result can comprise the question type of each question in the image to be corrected and the text area coordinates of each question. Referring to fig. 6, the inscription may include horizontal, vertical, and detached.
Fig. 7 is a diagram illustrating layout analysis results of a topic determination method according to another embodiment of the present application. The rectangular text box of the heavy thick line in fig. 7 represents a text region of each question in the layout analysis result. The rectangular text box with light-colored thin lines in fig. 7 represents the text area of each text line in the image to be corrected, which is identified by the centret model, that is, the text line detection result.
Referring to fig. 7, merging the layout analysis detection result and the text line detection result by using the intersection-to-union ratio may specifically include: and establishing a corresponding relation between the rectangular text boxes with dark thick lines and the rectangular text boxes with light thin lines belonging to the same topic. And analyzing to obtain the position relation of each question consisting of the text boxes and the question stem box and the answer box according to the characteristics of each question type in the layout analysis detection result. For example, for a cross-type question, it usually consists of a question stem box, an answer box and several text boxes for the intermediate solution process. And then, cutting the image to be corrected according to the corresponding relation, the text box composition characteristics of each question type and the coordinates of the text box forming the question, wherein the cut image comprises text areas of a question stem box and an answer box corresponding to each question in the image to be corrected.
Referring to fig. 3, in one embodiment, the method further comprises:
step S220, identifying the content of the text information in the question stem frame and the answer frame of the question;
step S230, obtaining at least one preliminary question judging result aiming at the question by utilizing at least one question judging model according to the result of content identification; and the number of the first and second groups,
in the case that all the preliminary answer results of the questions indicate that the answers to the questions are wrong, the step S120 of inputting the text information in the question stem frame to the question number processing model is performed.
In this embodiment, before the question number interference processing is performed, the question content recognition result is judged by using at least one question judgment model, and the question number interference processing is performed only when the question judgment result indicates that the answer to the question is wrong. Illustratively, in the photo-correction topic judgment strategy, there is a default rule of "judge the topic as correct as possible". Considering that most questions can be right when a normal user does questions, and only a few questions can be wrong, in the question judging strategy, the algorithm can generate a result which can judge the questions as correct as possible.
In general, if the content recognition result does not include the question number, the question model will give a correct question result. In this case, the subsequent header interference processing may not be performed. However, in the case where the question result indicates that the answer to the question is wrong, there may be a question number interference problem, and therefore, in this case, it is necessary to perform subsequent question number interference processing. The subsequent question mark interference processing procedure can comprise the following steps: inputting the text information in the question stem frame into a question number processing model, and outputting an adjusting mode and an adjusting proportion of the question stem frame by using the question number processing model; and adjusting the question stem frame according to the adjusting mode and the adjusting proportion.
In one example, a first problem assessment may be performed first using a first problem assessment model. In the first question judging process, the question number is not considered to be processed, and the question judging result of the question with question number interference may be a question judging result indicating that the answer of the question is wrong. And aiming at the question with wrong answer indicated by the question judging result, performing secondary question judgment by using a second question judging model. And performing third question judgment on the question which is judged to be wrong in the second question judgment or the question judged to be wrong in answer. And in the process of judging the question for the third time, executing a question number interference processing process.
In one embodiment, the first and second problem models may be different models. In another embodiment, the first and second problem models may be the same model, but the two models use different decoding methods, for example, Greedy Search (Greedy Search) or Beam Search (Beam Search) may be used as the decoding methods in the two models, respectively.
In another example, one problem evaluation model can be used for one problem evaluation to obtain one problem evaluation result. In the process of judging questions, the question numbers are not considered to be processed, and the question judging result of the question with question number interference may be a question judging result indicating that the answer of the question is wrong. And executing the question number interference processing process under the condition that the question judgment result is wrong.
In the embodiment, the problem number interference processing process is executed only for the problem with wrong answer indicated by the problem judgment result, so that the execution efficiency of the problem judgment method can be improved, and the problem judgment speed is increased.
In one embodiment, the method further comprises:
and training a second branch of the question mark processing model by adopting an intersection-to-parallel ratio loss function.
In one example, the first branch of the question mark processing model may be trained using a multi-class cross-entropy loss function. The cross entropy can be used to measure the difference degree of two different probability distributions in the same random variable, and can be used to measure the difference between the true probability distribution and the predicted probability distribution in machine learning. The smaller the value of the cross entropy, the better the model prediction effect.
In yet another example, the second branch of the topic processing model may be trained using a Dice Loss, i.e., cross-over-Loss function. In the model training process, the intersection and union ratio of the question stem frame obtained by the target detection model and the real question stem frame is calculated, namely the intersection and union ratio, and the intersection and union ratio is used as the value of the loss function.
FIG. 8 is a flow chart of model training for a topic determination method according to another embodiment of the present application. As shown in fig. 8, in one embodiment, the method further comprises:
step S310, adjusting the height coordinate value of the stem frame obtained by the target detection model to be the height coordinate value of the real stem frame;
and step S320, calculating the intersection ratio of the question stem frame obtained by the target detection model and the real question stem frame.
In order to eliminate the problem of question mark interference, the upper frame and the lower frame of the question stem frame do not need to be adjusted, and only the left frame and/or the right frame of the question stem frame need to be adjusted. When calculating the intersection ratio between the stem frame obtained by the target detection model and the real stem frame, first, the height coordinate value of the stem frame obtained by the target detection model is adjusted to be consistent with the height coordinate of the real stem frame, for example, the y-axis coordinate values of the upper frame and the lower frame of the two frames are adjusted to be consistent, and then the intersection ratio of the two frames is calculated. The processing mode only focuses on left and right changes of the question stem frame, so that the processing result can be more accurately focused on a task of calculating the loss value, and a better model training effect can be obtained.
Fig. 9 is a flowchart of a topic determination method according to another embodiment of the present application. As shown in fig. 9, an exemplary topic determination method comprises the following steps:
step 9.1: and constructing a training data set of the text line detection model.
Step 9.2: and training a text line detection model. For example, centret can be employed as a text line detection model.
Step 9.3: and constructing a question mark processing model. The question mark processing model may consist of two branches. The specific composition structure of each branch can be referred to the description in the foregoing, and is not repeated herein.
Step 9.4: and training the question number processing model according to the result output by the text line detection model. The first branch of the question mark processing model can be trained by adopting a multi-classification cross entropy loss function. The second branch of the question mark processing model can be trained by using an intersection-to-parallel ratio loss function.
Step 9.5: and shooting the image to be corrected, sending the image to be corrected into a shooting correction application program, and carrying out shooting question judgment operation according to a normal shooting question judgment flow.
Step 9.6: and (4) parallelly passing the image to be corrected sent in the step 9.5 through a layout analysis model, a text line detection model and a question stem answer detection model to obtain a layout analysis model detection result, a text line detection model detection result and a question stem answer detection model detection result. At least one of the layout analysis model and the question stem answer detection model can be selected. And then, according to the detection result, carrying out content identification by using an identification model to obtain a content identification result. And then, according to the content identification result, performing first question judgment. No question number is considered in the first question judgment, so that the question with question number interference can be judged wrongly. And the item with wrong judgment enters the second time of judging the item. The questions with question number interference in the second question judgment can be judged wrongly. And (4) aiming at the question which is judged wrongly in the second time of judging the question, entering a third time of judging the question.
Step 9.7: and for the questions entering the third question judgment, processing question number interference by using a question number processing model. For each question, whether horizontal, vertical or off-line, the text line detection model will detect its question stem information, at least one of the layout analysis model and question stem answer detection model will also detect its question stem information, and then the detected information is sent to the question number processing model to obtain the adjustment mode and the adjustment proportion. And obtaining a new question stem frame according to the adjustment mode and the adjustment proportion.
Step 9.8: and re-inputting the text surrounded by the new question stem box into the recognition model for content recognition.
Step 9.9: and judging the question according to the content identification result.
The problem of problem number interference is solved through the steps, and a more accurate problem judgment result is obtained.
Fig. 10 is a schematic structural diagram of a topic determination device according to an embodiment of the application. As shown in fig. 10, the apparatus may include:
the detection unit 100 is configured to perform text line detection on the image to be corrected to obtain a question stem frame of a question in the image to be corrected;
the processing unit 200 is used for inputting the text information in the question stem frame into the question mark processing model, processing the question mark in the question stem frame by using the question mark processing model, and outputting the adjustment mode and the adjustment proportion of the question stem frame;
the adjusting unit 300 is configured to adjust the stem frame according to the adjustment mode and the adjustment ratio to obtain an adjusted stem frame;
and the question judging unit 400 is used for obtaining a question judging result aiming at the question in the image to be corrected based on the adjusted question stem frame.
In one embodiment, the processing unit 200 is configured to:
outputting an adjusting mode of the question stem frame by utilizing a first branch of the question mark processing model; the adjusting mode comprises rightward adjustment of the left frame, leftward adjustment or no adjustment of the right frame.
In one embodiment, the processing unit 200 is configured to:
and outputting the adjustment proportion of the question stem frame by using the second branch of the question mark processing model.
In one embodiment, the detection unit 100 is further configured to:
and performing text line detection on the image to be corrected by using the target detection model to obtain a question stem frame and an answer frame of the question in the image to be corrected.
Fig. 11 is a schematic structural diagram of a topic determination device according to another embodiment of the present application. As shown in fig. 11, in one embodiment, the apparatus further includes a preliminary decision unit 150;
the preliminary topic determination unit 150 is configured to: identifying the content of the text information in the question stem frame and the answer frame of the question; obtaining at least one preliminary question judging result aiming at the question by utilizing at least one question judging model according to the result of the content identification;
the processing unit 200 is further configured to: and under the condition that all the preliminary question judging results of the questions indicate that the answers of the questions are wrong, executing the step of inputting the text information in the question stem frame into the question number processing model.
Fig. 12 is a schematic structural diagram of a topic determination device according to another embodiment of the present application. As shown in fig. 12, in one embodiment, the apparatus further comprises a training unit 250; the training unit 250 is configured to:
and training a second branch of the question mark processing model by adopting an intersection-to-parallel ratio loss function.
In one embodiment, the training unit 250 is further configured to:
adjusting the height coordinate value of the stem frame obtained by the target detection model into the height coordinate value of the real stem frame;
and calculating the intersection ratio of the question stem frame obtained by the target detection model and the real question stem frame.
The functions of each module in each apparatus in the embodiment of the present application may refer to corresponding descriptions in the above method, and are not described herein again.
FIG. 13 is a block diagram of an electronic device used to implement embodiments of the present application. As shown in fig. 13, the electronic apparatus includes: a memory 910 and a processor 920, the memory 910 having stored therein computer programs operable on the processor 920. The processor 920 implements the question determination and image generation method in the above-described embodiments when executing the computer program. The number of the memory 910 and the processor 920 may be one or more.
The electronic device further includes:
and a communication interface 930 for communicating with an external device to perform data interactive transmission.
If the memory 910, the processor 920 and the communication interface 930 are implemented independently, the memory 910, the processor 920 and the communication interface 930 may be connected to each other through a bus and perform communication with each other. The bus may be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus, an Extended ISA (EISA) bus, or the like. The bus may be divided into an address bus, a data bus, a control bus, etc. For ease of illustration, only one thick line is shown in FIG. 13, but this is not intended to represent only one bus or type of bus.
Optionally, in an implementation, if the memory 910, the processor 920 and the communication interface 930 are integrated on a chip, the memory 910, the processor 920 and the communication interface 930 may complete communication with each other through an internal interface.
Embodiments of the present application provide a computer-readable storage medium, which stores a computer program, and when the program is executed by a processor, the computer program implements the method provided in the embodiments of the present application.
The embodiment of the present application further provides a chip, where the chip includes a processor, and is configured to call and execute the instruction stored in the memory from the memory, so that the communication device in which the chip is installed executes the method provided in the embodiment of the present application.
An embodiment of the present application further provides a chip, including: the system comprises an input interface, an output interface, a processor and a memory, wherein the input interface, the output interface, the processor and the memory are connected through an internal connection path, the processor is used for executing codes in the memory, and when the codes are executed, the processor is used for executing the method provided by the embodiment of the application.
It should be understood that the processor may be a Central Processing Unit (CPU), other general purpose processor, a Digital Signal Processor (DSP), an Application Specific Integrated Circuit (ASIC), a Field Programmable Gate Array (FPGA) or other programmable logic device, discrete gate or transistor logic, discrete hardware components, etc. A general purpose processor may be a microprocessor or any conventional processor or the like. It is noted that the processor may be an advanced reduced instruction set machine (ARM) architecture supported processor.
Further, optionally, the memory may include a read-only memory and a random access memory, and may further include a nonvolatile random access memory. The memory may be either volatile memory or nonvolatile memory, or may include both volatile and nonvolatile memory. The non-volatile memory may include a read-only memory (ROM), a Programmable ROM (PROM), an Erasable PROM (EPROM), an electrically Erasable EPROM (EEPROM), or a flash memory. Volatile memory can include Random Access Memory (RAM), which acts as external cache memory. By way of example, and not limitation, many forms of RAM are available. For example, Static Random Access Memory (SRAM), Dynamic Random Access Memory (DRAM), Synchronous Dynamic Random Access Memory (SDRAM), double data rate synchronous SDRAM (DDR SDRAM), Enhanced SDRAM (ESDRAM), synchlink DRAM (SLDRAM), and direct memory bus RAM (DR RAM).
In the above embodiments, the implementation may be wholly or partially realized by software, hardware, firmware, or any combination thereof. When implemented in software, may be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions. The procedures or functions according to the present application are generated in whole or in part when the computer program instructions are loaded and executed on a computer. The computer may be a general purpose computer, a special purpose computer, a network of computers, or other programmable device. The computer instructions may be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another computer-readable storage medium.
In the description herein, references to the description of the term "one embodiment," "some embodiments," "an example," "a specific example," or "some examples," etc., mean that a particular feature, structure, material, or characteristic described in connection with the embodiment or example is included in at least one embodiment or example of the application. Furthermore, the particular features, structures, materials, or characteristics described may be combined in any suitable manner in any one or more embodiments or examples. Furthermore, various embodiments or examples and features of different embodiments or examples described in this specification can be combined and combined by one skilled in the art without contradiction.
Furthermore, the terms "first", "second" and "first" are used for descriptive purposes only and are not to be construed as indicating or implying relative importance or implicitly indicating the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one such feature. In the description of the present application, "a plurality" means two or more unless specifically limited otherwise.
Any process or method descriptions in flow charts or otherwise described herein may be understood as representing modules, segments, or portions of code which include one or more executable instructions for implementing specific logical functions or steps of the process. And the scope of the preferred embodiments of the present application includes other implementations in which functions may be performed out of the order shown or discussed, including substantially concurrently or in reverse order, depending on the functionality involved.
The logic and/or steps represented in the flowcharts or otherwise described herein, e.g., an ordered listing of executable instructions that can be considered to implement logical functions, can be embodied in any computer-readable medium for use by or in connection with an instruction execution system, apparatus, or device, such as a computer-based system, processor-containing system, or other system that can fetch the instructions from the instruction execution system, apparatus, or device and execute the instructions.
It should be understood that portions of the present application may be implemented in hardware, software, firmware, or a combination thereof. In the above embodiments, the various steps or methods may be implemented in software or firmware stored in memory and executed by a suitable instruction execution system. All or part of the steps of the method of the above embodiments may be implemented by hardware that is configured to be instructed to perform the relevant steps by a program, which may be stored in a computer-readable storage medium, and which, when executed, includes one or a combination of the steps of the method embodiments.
In addition, functional units in the embodiments of the present application may be integrated into one processing module, or each unit may exist alone physically, or two or more units are integrated into one module. The integrated module can be realized in a hardware mode, and can also be realized in a software functional module mode. The integrated module may also be stored in a computer-readable storage medium if it is implemented in the form of a software functional module and sold or used as a separate product. The storage medium may be a read-only memory, a magnetic or optical disk, or the like.
The above description is only for the specific embodiments of the present application, but the scope of the present application is not limited thereto, and any person skilled in the art can easily conceive various changes or substitutions within the technical scope of the present application, and these should be covered by the scope of the present application. Therefore, the protection scope of the present application shall be subject to the protection scope of the claims.
Claims (16)
1. A method for determining a subject, comprising:
performing text line detection on an image to be corrected to obtain a question stem frame of a question in the image to be corrected;
inputting the text information in the question stem frame into a question number processing model, processing the question number in the question stem frame by using the question number processing model, and outputting an adjusting mode and an adjusting proportion of the question stem frame;
adjusting the question stem frame according to the adjusting mode and the adjusting proportion to obtain an adjusted question stem frame;
and obtaining a question judging result aiming at the question in the image to be corrected based on the adjusted question stem frame.
2. The method of claim 1, wherein the processing the question mark in the question stem frame by using the question mark processing model and outputting the adjustment mode and the adjustment ratio of the question stem frame comprises:
outputting the adjusting mode of the question stem frame by utilizing the first branch of the question mark processing model; the adjusting mode comprises rightward adjustment of the left frame, leftward adjustment or no adjustment of the right frame.
3. The method of claim 1, wherein the processing the question mark in the question stem frame by using the question mark processing model and outputting the adjustment mode and the adjustment ratio of the question stem frame comprises:
and outputting the adjustment proportion of the question stem frame by using the second branch of the question mark processing model.
4. The method according to any one of claims 1 to 3, wherein text line detection is performed on an image to be corrected to obtain a stem frame of a subject in the image to be corrected, and the method further comprises:
and performing text line detection on the image to be corrected by using the target detection model to obtain a question stem frame and an answer frame of the question in the image to be corrected.
5. The method of claim 4, further comprising:
identifying the content of the text information in the question stem frame and the answer frame of the question;
obtaining at least one preliminary question judging result aiming at the question by utilizing at least one question judging model according to the content identification result;
and under the condition that all the preliminary question judging results of the questions indicate that the answers of the questions are wrong, executing the step of inputting the text information in the question stem frame into the question number processing model.
6. The method of claim 4, further comprising:
and training a second branch of the question mark processing model by adopting an intersection ratio loss function.
7. The method of claim 6, further comprising:
adjusting the height coordinate value of the question stem frame obtained by the target detection model into the height coordinate value of the real question stem frame;
and calculating the intersection ratio of the question stem frame obtained by the target detection model and the real question stem frame.
8. A question determination apparatus, comprising:
the detection unit is used for detecting text lines of an image to be corrected to obtain a question stem frame of a question in the image to be corrected;
the processing unit is used for inputting the text information in the question stem frame into a question mark processing model, processing the question mark in the question stem frame by using the question mark processing model and outputting the adjustment mode and the adjustment proportion of the question stem frame;
the adjusting unit is used for adjusting the question stem frame according to the adjusting mode and the adjusting proportion to obtain an adjusted question stem frame;
and the question judging unit is used for obtaining a question judging result aiming at the question in the image to be corrected based on the adjusted question stem frame.
9. The apparatus of claim 8, wherein the processing unit is configured to:
outputting the adjusting mode of the question stem frame by utilizing the first branch of the question mark processing model; the adjusting mode comprises rightward adjustment of the left frame, leftward adjustment or no adjustment of the right frame.
10. The apparatus of claim 8, wherein the processing unit is configured to:
and outputting the adjustment proportion of the question stem frame by using the second branch of the question mark processing model.
11. The apparatus of any one of claims 8 to 10, wherein the detection unit is further configured to:
and performing text line detection on the image to be corrected by using the target detection model to obtain a question stem frame and an answer frame of the question in the image to be corrected.
12. The apparatus according to claim 11, further comprising a preliminary topic determination unit;
the preliminary judgment unit is used for: identifying the content of the text information in the question stem frame and the answer frame of the question; obtaining at least one preliminary question judging result aiming at the question by utilizing at least one question judging model according to the content identification result;
the processing unit is further to: and under the condition that all the preliminary question judging results of the questions indicate that the answers of the questions are wrong, executing the step of inputting the text information in the question stem frame into the question number processing model.
13. The apparatus of claim 11, further comprising a training unit; the training unit is configured to:
and training a second branch of the question mark processing model by adopting an intersection ratio loss function.
14. The apparatus of claim 13, wherein the training unit is further configured to:
adjusting the height coordinate value of the question stem frame obtained by the target detection model into the height coordinate value of the real question stem frame;
and calculating the intersection ratio of the question stem frame obtained by the target detection model and the real question stem frame.
15. An electronic device comprising a processor and a memory, the memory having stored therein instructions that are loaded and executed by the processor to implement the method of any of claims 1 to 7.
16. A computer-readable storage medium, in which a computer program is stored which, when being executed by a processor, carries out the method according to any one of claims 1 to 7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110005089.0A CN112347998A (en) | 2021-01-05 | 2021-01-05 | Question judging method, device, equipment and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110005089.0A CN112347998A (en) | 2021-01-05 | 2021-01-05 | Question judging method, device, equipment and storage medium |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN112347998A true CN112347998A (en) | 2021-02-09 |
Family
ID=74427757
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110005089.0A Pending CN112347998A (en) | 2021-01-05 | 2021-01-05 | Question judging method, device, equipment and storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112347998A (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112990180A (en) * | 2021-04-29 | 2021-06-18 | 北京世纪好未来教育科技有限公司 | Question judging method, device, equipment and storage medium |
| CN112990102A (en) * | 2021-04-16 | 2021-06-18 | 四川阿泰因机器人智能装备有限公司 | Improved Centernet complex environment target detection method |
| CN113239909A (en) * | 2021-07-12 | 2021-08-10 | 北京世纪好未来教育科技有限公司 | Question processing method, device, equipment and medium |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20170262738A1 (en) * | 2014-09-16 | 2017-09-14 | Iflytek Co., Ltd. | Intelligent scoring method and system for text objective question |
| CN111753120A (en) * | 2020-06-29 | 2020-10-09 | 广东小天才科技有限公司 | Method and device for searching questions, electronic equipment and storage medium |
| CN111767830A (en) * | 2020-06-28 | 2020-10-13 | 广东小天才科技有限公司 | Page scanning frame question method, system, terminal equipment and storage medium |
| CN111931730A (en) * | 2020-09-24 | 2020-11-13 | 北京易真学思教育科技有限公司 | Question judging method and device, electronic equipment and storage medium |
| CN111931731A (en) * | 2020-09-24 | 2020-11-13 | 北京易真学思教育科技有限公司 | Question judging method and device, electronic equipment and storage medium |
-
2021
- 2021-01-05 CN CN202110005089.0A patent/CN112347998A/en active Pending
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20170262738A1 (en) * | 2014-09-16 | 2017-09-14 | Iflytek Co., Ltd. | Intelligent scoring method and system for text objective question |
| CN111767830A (en) * | 2020-06-28 | 2020-10-13 | 广东小天才科技有限公司 | Page scanning frame question method, system, terminal equipment and storage medium |
| CN111753120A (en) * | 2020-06-29 | 2020-10-09 | 广东小天才科技有限公司 | Method and device for searching questions, electronic equipment and storage medium |
| CN111931730A (en) * | 2020-09-24 | 2020-11-13 | 北京易真学思教育科技有限公司 | Question judging method and device, electronic equipment and storage medium |
| CN111931731A (en) * | 2020-09-24 | 2020-11-13 | 北京易真学思教育科技有限公司 | Question judging method and device, electronic equipment and storage medium |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112990102A (en) * | 2021-04-16 | 2021-06-18 | 四川阿泰因机器人智能装备有限公司 | Improved Centernet complex environment target detection method |
| CN112990180A (en) * | 2021-04-29 | 2021-06-18 | 北京世纪好未来教育科技有限公司 | Question judging method, device, equipment and storage medium |
| CN112990180B (en) * | 2021-04-29 | 2021-08-06 | 北京世纪好未来教育科技有限公司 | Question judging method, device, equipment and storage medium |
| CN113239909A (en) * | 2021-07-12 | 2021-08-10 | 北京世纪好未来教育科技有限公司 | Question processing method, device, equipment and medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN112347998A (en) | Question judging method, device, equipment and storage medium | |
| CN111931731B (en) | Question judging method and device, electronic equipment and storage medium | |
| CN111931730B (en) | Question judging method and device, electronic equipment and storage medium | |
| CN108885699A (en) | Character identifying method, device, storage medium and electronic equipment | |
| US11042742B1 (en) | Apparatus and method for detecting road based on convolutional neural network | |
| CN112990180B (en) | Question judging method, device, equipment and storage medium | |
| CN111814902A (en) | Target detection model training method, target identification method, device and medium | |
| CN112418278A (en) | Multi-class object detection method, terminal device and storage medium | |
| CN111738249B (en) | Image detection method, image detection device, electronic equipment and storage medium | |
| CN111737478B (en) | Text detection method, electronic device and computer readable medium | |
| CN111985469B (en) | Method and device for recognizing characters in image and electronic equipment | |
| CN112257703B (en) | Image recognition method, device, equipment and readable storage medium | |
| CN111985465A (en) | Text recognition method, device, equipment and storage medium | |
| CN109389110B (en) | Region determination method and device | |
| CN112989768B (en) | Method and device for correcting connection questions, electronic equipment and storage medium | |
| CN113112511B (en) | Method and device for correcting test paper, storage medium and electronic equipment | |
| CN112308053A (en) | Detection model training and question judging method and device, electronic equipment and storage medium | |
| CN113239909B (en) | Question processing method, device, equipment and medium | |
| CN113807407B (en) | Target detection model training method, model performance detection method and device | |
| CN114445716A (en) | Key point detection method, key point detection device, computer device, medium, and program product | |
| CN113392455A (en) | House type graph scale detection method and device based on deep learning and electronic equipment | |
| CN113315995A (en) | Method and device for improving video quality, readable storage medium and electronic equipment | |
| CN112052883A (en) | Clothes detection method, device and storage medium | |
| CN109214230B (en) | Data matrix code identification method and device and electronic equipment | |
| CN117474932B (en) | Object segmentation method and device, electronic equipment and storage medium |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20210209 |
|
| RJ01 | Rejection of invention patent application after publication |