Door notice detection method and device, medium and equipment

Abstract

The invention provides an on-door notification bill detection method and device, a medium and equipment. The method comprises the following steps: determining a to-be-detected picture; wherein the to-be-detected picture is a picture obtained by photographing a door; outputting the to-be-detected picture to a pre-trained target detection model to obtain a detection result; wherein the detection result comprises the position of the detection target of which the category is a door and the position of the detection target of which the category is a candidate notice; according to the position of the detection target of which the category is the door and the position of the detection target of which the category is the candidate notice, determining whether the candidate notice is attached to the door; if it is determined that the candidate notification bill is pasted on the door, whether characters of a'notification bill 'exist in the candidate notification bill or not is recognized; and if the characters of the'notification bill 'exist, determining that the notification bill is detected on the door, and taking the candidate notification bill as the detected notification bill. According to the invention, the detection accuracy of the notice on the door can be obviously improved.

Description

Technical Field

[0001] The present invention relates to a target technical field, and in particular to a door notice detection method, device, medium and equipment. Background Art

[0002] Object detection is an important task in computer vision, which aims to accurately identify and locate the target object of interest from an image or video. Traditional object detection methods mainly rely on manually designed features and classifiers, but this method does not work well when dealing with complex scenes and diverse targets. For example, in the scene of detecting notices on the door, because there may be advertisements and other notices similar to notices posted on the door, it is more difficult to detect the notices from these notices, so it is necessary to provide a detection solution with higher accuracy. Summary of the invention

[0003] In view of at least one of the above technical problems, an embodiment of the present invention provides a method and device, medium, and equipment for detecting a door notice.

[0004] According to the first aspect, the door notice detection method provided by the embodiment of the present invention includes:

[0005] Determine a picture to be detected; wherein the picture to be detected is a picture obtained by taking a photo of the door;

[0006] Outputting the image to be detected to a pre-trained target detection model to obtain a detection result; wherein the detection result includes the position of the detection target of the door category and the position of the detection target of the candidate notice category;

[0007] Determine whether the candidate notice is posted on the door according to the position of the detection target of the door category and the position of the detection target of the candidate notice category;

[0008] If it is determined that the candidate notice is posted on the door, identifying whether there is a word "notice" in the candidate notice;

[0009] If the word "notice" exists, it is determined that a notice is detected on the door, and the candidate notice is used as the detected notice.

[0010] In one embodiment, the detection result also includes the confidence level corresponding to the door and the confidence level corresponding to the candidate notification form;

[0011] Correspondingly, before determining whether the candidate notification is posted on the door according to the position of the detection target of the door category and the position of the detection target of the candidate notification category, the method further includes:

[0012] Determine whether the confidence corresponding to the door and the confidence corresponding to the candidate notification are both higher than a preset confidence;

[0013] If so, the step of determining whether the candidate notice is posted on the door according to the position of the detection target whose category is door and the position of the detection target whose category is candidate notice is executed.

[0014] In one embodiment, the method further includes: if the confidence level corresponding to the door is lower than the preset confidence level, prompting information that the image to be detected has errors.

[0015] In one embodiment, the method further includes: if the confidence level corresponding to the candidate notification is lower than the preset confidence level, prompting information indicating that the image to be detected has no notification to be detected.

[0016] In one embodiment, the identifying whether the candidate notifications contain the word “notification” includes: identifying whether the candidate notifications contain the word “notification” by using optical character recognition technology.

[0017] In one embodiment, the position of the detection target in the output result is the coordinate information of four vertices of the detection frame of the detection target;

[0018] Correspondingly, determining whether the candidate notification is posted on the door according to the position of the detection target of the door category and the position of the detection target of the candidate notification category includes:

[0019] Calculate the intersection-over-union ratio between the detection frame of the detection target of the category being the door and the detection frame of the detection target of the category being the candidate notification form according to the coordinate information of the four vertices of the detection frame of the detection target of the category being the door and the coordinate information of the four vertices of the detection frame of the detection target of the category being the candidate notification form;

[0020] Determining whether the intersection-over-union ratio is greater than a preset intersection-over-union ratio;

[0021] If so, confirm that the candidate notice is posted on the door;

[0022] Otherwise, it is determined that the candidate notice is not posted on the door.

[0023] In one embodiment, the loss function used in the training process of the target detection model is y=1-x, where x is the intersection-and-union ratio between the position of the detection target marked in the training sample and the position of the detection target output by the model, and y is the loss value; and / or, during the training process of the target detection model, a preset learning rate decay period and decay ratio are set to provide model stability and accuracy.

[0024] According to a second aspect, a door notice detection device provided by an embodiment of the present invention includes:

[0025] A picture determination module is used to determine a picture to be detected; wherein the picture to be detected is a picture obtained by taking a photo of the door;

[0026] A model detection module, used to output the image to be detected to a pre-trained target detection model to obtain a detection result; wherein the detection result includes the position of the detection target of the door category and the position of the detection target of the candidate notice category;

[0027] A first determination module, configured to determine whether the candidate notice is posted on the door according to the position of the detection target of the door category and the position of the detection target of the candidate notice category;

[0028] A text recognition module, for identifying whether the candidate notice contains the word "notice" if it is determined that the candidate notice is posted on the door;

[0029] The result determination module is used to determine that a notice is detected on the door if there is a word "notice", and to use the candidate notice as the detected notice.

[0030] In one embodiment, the detection result also includes the confidence corresponding to the door and the confidence corresponding to the candidate notification; correspondingly, before the model detection module determines whether the candidate notification is posted on the door based on the position of the detection target classified as the door and the position of the detection target classified as the candidate notification, it is also used to: determine whether the confidence corresponding to the door and the confidence corresponding to the candidate notification are both higher than a preset confidence; if so, execute the method of determining whether the candidate notification is posted on the door based on the position of the detection target classified as the door and the position of the detection target classified as the candidate notification.

[0031] In one embodiment, the model detection module is further used to: if the confidence corresponding to the door is lower than the preset confidence, then prompt information that the image to be detected has errors.

[0032] In one embodiment, the model detection module is further used to: if the confidence corresponding to the candidate notification is lower than the preset confidence, prompt information that the image to be detected does not have a notification to be detected.

[0033] In one embodiment, the text recognition module is specifically used to: identify whether the text "notification" exists in the candidate notifications through optical character recognition technology.

[0034] In one embodiment, the position of the detection target in the output result is the coordinate information of the four vertices of the detection frame of the detection target; correspondingly, the first determination module is specifically used to: calculate the intersection-and-union ratio between the detection frame of the detection target whose category is the door and the detection frame of the detection target whose category is the candidate notification according to the coordinate information of the four vertices of the detection frame of the detection target whose category is the door and the coordinate information of the four vertices of the detection frame of the detection target whose category is the candidate notification; determine whether the intersection-and-union ratio is greater than a preset intersection-and-union ratio; if so, determine that the candidate notification is affixed to the door; otherwise, determine that the candidate notification is not affixed to the door.

[0035] In one embodiment, the apparatus further comprises:

[0036] A model training module is used to adopt a loss function of y=1-x during the training process of the target detection model, where x is the intersection-and-union ratio between the position of the detection target marked in the training sample and the position of the detection target output by the model, and y is the loss value; and / or, during the training process of the target detection model, provide model stability and accuracy by setting a preset learning rate decay period and decay ratio.

[0037] According to a third aspect, an embodiment of the present invention provides a computer-readable storage medium having a computer program stored thereon, which, when executed in a computer, enables the computer to execute the method provided in the first aspect.

[0038] According to a fourth aspect, an embodiment of the present invention provides a computing device, including a memory and a processor, wherein the memory stores executable code, and when the processor executes the executable code, the method provided in the first aspect is implemented.

[0039] The door notice detection method, device, medium, and equipment provided by the embodiment of the present invention output the image to be detected to the pre-trained target detection model, obtain the position of the detection target of the door category and the position of the detection target of the candidate notice category, and then determine whether the candidate notice is posted on the door; if the candidate notice is posted on the door, then identify whether the text of "notice" exists in the candidate notice, and if the text of "notice" exists, then determine that the notice is detected on the door. It can be seen that the embodiment of the present invention uses a model to identify the relevant information of the door and the candidate notice, and then determines whether the candidate notice is posted on the door, and then identifies whether the notice text exists. It can be seen that the embodiment of the present invention integrates multiple information to distinguish the notice on the door from other orders such as advertisements, which can significantly improve the detection accuracy of the door notice. Since the entire detection process does not require human participation, the detection speed and efficiency can be greatly improved. It can be seen that the embodiment of the present invention realizes accurate detection of small targets based on the target detection model, and is suitable for fast and accurate target detection in complex scenes. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] Figure 1 It is a flowchart of a method for detecting a door-to-door notification in one embodiment of the present invention;

[0041] Figure 2 It is a structural block diagram of a door notice detection device in one embodiment of the present invention. DETAILED DESCRIPTION

[0042] In a first aspect, an embodiment of the present invention provides a method for detecting a door notification. Figure 1 The method comprises the following steps S110 to S150:

[0043] S110, determining a picture to be detected; wherein the picture to be detected is a picture obtained by taking a photo of the door;

[0044] For example, take a picture of the door through the camera at the door, and then obtain the captured picture from the camera.

[0045] It is understandable that in the image obtained by taking a photo of a door, there may be some areas that do not belong to the door, such as the door frame, the wall where the door frame is located, etc. In other words, the image to be detected contains not only the door, but also parts other than the door are usually photographed.

[0046] S120, outputting the image to be detected to a pre-trained target detection model to obtain a detection result; wherein the detection result includes the position of the detection target of the door category and the position of the detection target of the candidate notice category;

[0047] That is, when the image to be detected is input into the target detection model, the target detection model will output information such as category, location, and even the confidence mentioned below. The category refers to whether it is a door or a candidate notification, the location of the door refers to the coordinate information of the four vertices of the door's detection frame, and the location of the candidate notification refers to the coordinate information of the four vertices of the candidate notification's detection frame.

[0048] The candidate notice refers to the notice identified from the image to be detected. The notice is most likely a notice, but it may also be other types of notices, such as inspection notices, advertisements, etc. Moreover, the candidate notice may be posted on the door, on the door frame, or on the wall.

[0049] S130, determining whether the candidate notification sheet is posted on the door according to the position of the detection target of the door category and the position of the detection target of the candidate notification sheet category;

[0050] That is to say, whether the candidate notice is posted on the door is determined based on the positions of the two detection boxes output by the target detection model.

[0051] In one embodiment, the position of the detection target in the output result is the coordinate information of the four vertices of the detection frame of the detection target; correspondingly, determining whether the candidate notice is posted on the door according to the position of the detection target whose category is door and the position of the detection target whose category is candidate notice in S130 may include S131 to S134:

[0052] S131, calculating the intersection-over-union ratio between the detection frame of the detection target of the category being the door and the detection frame of the detection target of the category being the candidate notification form according to the coordinate information of the four vertices of the detection frame of the detection target of the category being the door and the coordinate information of the four vertices of the detection frame of the detection target of the category being the candidate notification form;

[0053] S132, determining whether the intersection-over-combination ratio is greater than a preset intersection-over-combination ratio;

[0054] S133, if yes, determine that the candidate notice is posted on the door;

[0055] S134. Otherwise, determine that the candidate notice is not posted on the door.

[0056] Among them, the intersection-to-union ratio refers to the ratio of the intersection of two detection frames to their union. It can be seen that the larger the intersection of the two detection frames, the larger the intersection-to-union ratio, and the greater the degree of intersection between the gate and the candidate notification.

[0057] The preset intersection-and-union ratio can be set as needed, for example, set to 0.6, that is, as long as the intersection-and-union ratio is greater than 0.6, it is considered that the candidate notice is posted on the door.

[0058] It can be seen that determining the positional relationship between the candidate notification and the door by calculating the intersection-and-union ratio is very simple, fast and accurate.

[0059] S140, if it is determined that the candidate notice is posted on the door, identifying whether there is a word "notice" in the candidate notice;

[0060] It is understandable that the notice will generally have the word "notice" on it, for example, "Water Bill Payment Notice", "Corridor Debris Disposal Notice", etc.

[0061] In one embodiment, the identifying whether the word “notice” exists in the candidate notification in S140 may include: identifying whether the word “notice” exists in the candidate notification by using optical character recognition technology.

[0062] Among them, optical character recognition technology is OCR technology.

[0063] Specifically, OCR technology is used to grayscale the image, locate text, segment characters, extract text on the notice, and determine whether the word "notice" exists. It can be seen that based on OCR technology, text recognition is performed on the detected candidate notices to detect whether the candidate notices contain the "notice" field to distinguish between notices and other forms such as advertisements.

[0064] S150. If the word "notice" exists, it is determined that a notice is detected on the door, and the candidate notice is used as the detected notice.

[0065] It can be seen that only when the candidate notice is posted on the door and the candidate notice contains the word "notice", it is determined that the notice is detected on the door. However, if the candidate notice is not posted on the door and / or the candidate notice does not contain the word "notice", it is considered that the notice is not detected on the door.

[0066] In one embodiment, the detection result also includes the confidence level corresponding to the door and the confidence level corresponding to the candidate notification form;

[0067] Correspondingly, before determining whether the candidate notification is posted on the door based on the position of the detection target classified as a door and the position of the detection target classified as the candidate notification in S120, the method may further include: judging whether the confidence corresponding to the door and the confidence corresponding to the candidate notification are both higher than a preset confidence; if so, executing the step of determining whether the candidate notification is posted on the door based on the position of the detection target classified as a door and the position of the detection target classified as the candidate notification.

[0068] That is to say, the results output by the model include two confidence levels, one confidence level is the probability that the detection box of the detection target with the category of door corresponds to the door, and the other confidence level is the probability that the detection box of the detection target with the category of candidate notice corresponds to the candidate notice.

[0069] It can be seen that only when the two confidences output by the model are higher than the preset confidences, S120 and subsequent steps are performed. Here, the accuracy of notification detection can be further improved by the confidence.

[0070] Furthermore, the method may also include: if the confidence level corresponding to the door is lower than the preset confidence level, prompting information that the image to be detected has errors.

[0071] That is to say, if the confidence corresponding to the door is lower than the preset confidence, it means that the picture to be detected is not a picture taken facing the door, and there is no door in the picture to be detected, so an error prompt message is output, so that relevant personnel can know that the reason for the non-detection notice is that there is no door in the picture to be detected.

[0072] Furthermore, the method may further include: if the confidence level corresponding to the candidate notification is lower than the preset confidence level, prompting information indicating that there is no notification to be detected in the image to be detected.

[0073] That is to say, if the confidence level corresponding to the candidate notification is lower than the preset confidence level, it means that there is no notification that needs to be detected in the image to be detected.

[0074] It is understandable that after a target detection model is input with a picture to be detected, it will identify multiple detection targets of the same category in the picture to be detected, that is, there are multiple detection frames for the same category of detection targets, each detection frame has a confidence level, and the detection frame with the highest confidence level is selected as the position output of the detection target of this category. If the confidence level corresponding to the detection target of this category is lower than the preset confidence level, it means that the confidence level is very low and the reliability is not high, so it cannot be adopted. At this time, it is considered that there is no notification that needs to be detected in the picture to be detected.

[0075] During the training of the target detection model, the offline images are first processed. For example, the images are randomly scaled, rotated, cropped, and other enhancement methods are used to obtain multiple images, which can increase the number and diversity of images, thereby improving the generalization ability of the trained model. Then the detection frame of the door and the detection frame of the notice are marked on the image. After marking, a picture is called a sample, and multiple samples form a sample set. The yolov8m framework is trained using the sample set, and the loss value is calculated after each training. Determine whether the loss value is lower than the preset value. If it is not lower than the preset value, continue training. If it is lower than the preset value, the model is considered to have converged.

[0076] In one embodiment, the loss function used in the training process of the target detection model can be y=1-x, where x is the intersection-and-union ratio between the position of the detection target marked in the training sample and the position of the detection target output by the model, and y is the loss value.

[0077] It is understandable that when a sample is input into the target detection model, the larger the intersection-and-union ratio between the output detection box and the true detection box marked on the sample, the closer the output detection box is to the true detection box, that is, the higher the accuracy of the model output, and the smaller the loss value.

[0078] Calculating the loss value using the formula y=1-x is not only simple, but also fast and accurate.

[0079] In one embodiment, during the training process of the target detection model, a preset learning rate decay period and decay ratio are set to provide model stability and accuracy.

[0080] For example, the preset learning rate decay cycle is set to 30 rounds, that is, 30 epochs decay once, and the decay ratio, that is, the decay rate, is 0.1. Epoch in a neural network refers to the process in which the entire sample set is completely traversed by the neural network. The advantage of doing this is that it can improve the accuracy and stability of the model.

[0081] The embodiment of the present invention aims at the detection demand of door notices. It uses a model to identify the relevant information of the door and the candidate notices, then determines whether the candidate notice is posted on the door, and uses optical character recognition technology to identify whether there is text on the notice. It can be seen that the embodiment of the present invention integrates a variety of information to distinguish the notice on the door from other forms such as advertisements, which can significantly improve the detection accuracy of the door notice. Since the entire detection process does not require human participation, the detection speed and efficiency can be greatly improved. It can be seen that the embodiment of the present invention realizes accurate detection of small targets based on the target detection model, and is suitable for fast and accurate target detection in complex scenarios.

[0082] In a second aspect, an embodiment of the present invention provides a door notice detection device, see Figure 2 , the device 100 comprises:

[0083] The picture determination module 110 is used to determine the picture to be detected; wherein the picture to be detected is a picture obtained by taking a photo of the door;

[0084] The model detection module 120 is used to output the image to be detected to the pre-trained target detection model to obtain a detection result; wherein the detection result includes the position of the detection target of the door category and the position of the detection target of the candidate notice category;

[0085] A first determination module 130, configured to determine whether the candidate notice is posted on the door according to the position of the detection target of the door category and the position of the detection target of the candidate notice category;

[0086] The text recognition module 140 is used to identify whether the text "notice" exists in the candidate notice if it is determined that the candidate notice is posted on the door;

[0087] The result determination module 150 is used to determine that a notice is detected on the door if there is a word "notice", and to use the candidate notice as the detected notice.

[0088] In one embodiment, the detection result also includes the confidence corresponding to the door and the confidence corresponding to the candidate notification; correspondingly, before the model detection module determines whether the candidate notification is posted on the door based on the position of the detection target classified as the door and the position of the detection target classified as the candidate notification, it is also used to: determine whether the confidence corresponding to the door and the confidence corresponding to the candidate notification are both higher than a preset confidence; if so, execute the method of determining whether the candidate notification is posted on the door based on the position of the detection target classified as the door and the position of the detection target classified as the candidate notification.

[0089] In one embodiment, the model detection module is further used to: if the confidence corresponding to the door is lower than the preset confidence, then prompt information that the image to be detected has errors.

[0090] In one embodiment, the model detection module is further used to: if the confidence corresponding to the candidate notification is lower than the preset confidence, prompt information that the image to be detected does not have a notification to be detected.

[0091] In one embodiment, the text recognition module is specifically used to: identify whether the text "notification" exists in the candidate notifications through optical character recognition technology.

[0092] In one embodiment, the position of the detection target in the output result is the coordinate information of the four vertices of the detection frame of the detection target; correspondingly, the first determination module is specifically used to: calculate the intersection-and-union ratio between the detection frame of the detection target whose category is the door and the detection frame of the detection target whose category is the candidate notification according to the coordinate information of the four vertices of the detection frame of the detection target whose category is the door and the coordinate information of the four vertices of the detection frame of the detection target whose category is the candidate notification; determine whether the intersection-and-union ratio is greater than a preset intersection-and-union ratio; if so, determine that the candidate notification is affixed to the door; otherwise, determine that the candidate notification is not affixed to the door.

[0093] In one embodiment, the apparatus further comprises:

[0094] A model training module is used to adopt a loss function of y=1-x during the training process of the target detection model, where x is the intersection-and-union ratio between the position of the detection target marked in the training sample and the position of the detection target output by the model, and y is the loss value; and / or, during the training process of the target detection model, provide model stability and accuracy by setting a preset learning rate decay period and decay ratio.

[0095] It is understandable that the explanation, specific implementation, beneficial effects, examples, etc. of the relevant contents in the device provided in the embodiment of the present invention can be found in the corresponding parts of the method provided in the first aspect, and will not be repeated here.

[0096] In a third aspect, an embodiment of the present invention provides a computer-readable medium, on which computer instructions are stored. When the computer instructions are executed by a processor, the processor executes the method provided in the first aspect.

[0097] Specifically, a system or device equipped with a storage medium can be provided, on which software program code that implements the functions of any of the above-mentioned embodiments is stored, and a computer (or CPU or MPU) of the system or device can be enabled to read and execute the program code stored in the storage medium.

[0098] In this case, the program code itself read from the storage medium can realize the function of any one of the above-mentioned embodiments, and thus the program code and the storage medium storing the program code constitute a part of the present invention.

[0099] The storage medium embodiments for providing the program code include a floppy disk, a hard disk, a magneto-optical disk, an optical disk (such as CD-ROM, CD-R, CD-RW, DVD-ROM, DVD-RAM, DVD-RW, DVD+RW), a magnetic tape, a non-volatile memory card, and a ROM. Alternatively, the program code can be downloaded from a server computer by a communication network.

[0100] In addition, it should be clear that the functions of any of the above embodiments can be implemented not only by executing the program code read by the computer, but also by enabling an operating system operating on the computer to complete part or all of the actual operations based on instructions from the program code.

[0101] In addition, it can be understood that the program code read from the storage medium is written to a memory provided in an expansion board inserted into the computer or to a memory provided in an expansion module connected to the computer, and then based on the instructions of the program code, a CPU installed on the expansion board or expansion module is enabled to perform part or all of the actual operations, thereby realizing the functions of any of the above-mentioned embodiments.

[0102] It is understandable that the explanation, specific implementation methods, beneficial effects, examples, etc. of the relevant contents in the computer-readable medium provided in the embodiment of the present invention can be found in the corresponding parts of the method provided in the first aspect, and will not be repeated here.

[0103] In a fourth aspect, an embodiment of the present specification provides a computing device, including a memory and a processor, wherein the memory stores executable code, and when the processor executes the executable code, it implements the method in any embodiment of the specification.

[0104] It is understandable that the explanation, specific implementation, beneficial effects, examples, etc. of the relevant contents in the computing device provided in the embodiment of the present invention can be found in the corresponding parts of the method provided in the first aspect, and will not be repeated here.

[0105] Each embodiment in this specification is described in a progressive manner, and the same or similar parts between the embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the device embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment.

[0106] Those skilled in the art should be aware that in one or more of the above examples, the functions described in the present invention can be implemented by hardware, software, widgets, or any combination thereof. When implemented by software, these functions can be stored in a computer-readable medium or transmitted as one or more instructions or codes on a computer-readable medium.

[0107] The specific implementation methods described above further illustrate the objectives, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above description is only a specific implementation method of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc. made on the basis of the technical solution of the present invention should be included in the scope of protection of the present invention.

Claims

1. A door notice detection method, characterized in that: include: Determine a picture to be detected; wherein the picture to be detected is a picture obtained by taking a photo of the door; Outputting the image to be detected to a pre-trained target detection model to obtain a detection result; wherein the detection result includes the position of the detection target of the door category and the position of the detection target of the candidate notice category; Determine whether the candidate notice is posted on the door according to the position of the detection target of the door category and the position of the detection target of the candidate notice category; If it is determined that the candidate notice is posted on the door, identifying whether there is a word "notice" in the candidate notice; If the word "notice" exists, it is determined that a notice is detected on the door, and the candidate notice is used as the detected notice.

2. The method according to claim 1, characterized in that The detection result also includes the confidence level corresponding to the door and the confidence level corresponding to the candidate notification form; Correspondingly, before determining whether the candidate notification is posted on the door according to the position of the detection target of the door category and the position of the detection target of the candidate notification category, the method further includes: Determine whether the confidence corresponding to the door and the confidence corresponding to the candidate notification are both higher than a preset confidence; If so, the step of determining whether the candidate notice is posted on the door according to the position of the detection target whose category is door and the position of the detection target whose category is candidate notice is executed.

3. The method according to claim 2, characterized in that Also includes: If the confidence level corresponding to the gate is lower than the preset confidence level, it is prompted that there is an error in the image to be detected.

4. The method according to claim 2, characterized in that: Also includes: If the confidence level corresponding to the candidate notification is lower than the preset confidence level, it is prompted that there is no notification level to be detected in the image to be detected.

5. The method according to claim 1, characterized in that The step of identifying whether the candidate notification contains the word "notification", comprises: By using optical character recognition technology, it is determined whether the word "notice" exists in the candidate notice.

6. The method according to claim 1, characterized in that The position of the detection target in the output result is the coordinate information of the four vertices of the detection frame of the detection target; Correspondingly, determining whether the candidate notification is posted on the door according to the position of the detection target of the door category and the position of the detection target of the candidate notification category includes: Calculate the intersection-over-union ratio between the detection frame of the detection target of the category being the door and the detection frame of the detection target of the category being the candidate notification form according to the coordinate information of the four vertices of the detection frame of the detection target of the category being the door and the coordinate information of the four vertices of the detection frame of the detection target of the category being the candidate notification form; Determining whether the intersection-over-union ratio is greater than a preset intersection-over-union ratio; If so, confirm that the candidate notice is posted on the door; Otherwise, it is determined that the candidate notice is not posted on the door.

7. The method according to claim 1, characterized in that The loss function used in the training process of the target detection model is y=1-x, where x is the intersection-and-union ratio between the position of the detection target marked in the training sample and the position of the detection target output by the model, and y is the loss value; and / or, during the training process of the target detection model, a preset learning rate decay period and decay ratio are set to provide model stability and accuracy.

8. A door notice detection device, characterized in that: include: A picture determination module is used to determine a picture to be detected; wherein the picture to be detected is a picture obtained by taking a photo of the door; A model detection module, used to output the image to be detected to a pre-trained target detection model to obtain a detection result; wherein the detection result includes the position of the detection target of the door category and the position of the detection target of the candidate notice category; A first determination module, configured to determine whether the candidate notice is posted on the door according to the position of the detection target of the door category and the position of the detection target of the candidate notice category; A text recognition module, for identifying whether the candidate notice contains the word "notice" if it is determined that the candidate notice is posted on the door; The result determination module is used to determine that a notice is detected on the door if the word "notice" exists, and to use the candidate notice as the detected notice.

9. A computer-readable storage medium, characterized in that: A computer program is stored thereon, and when the computer program is executed in a computer, the computer is caused to execute the method according to any one of claims 1 to 7.

10. A computing device, characterized in that: The method comprises a memory and a processor, wherein the memory stores executable codes, and when the processor executes the executable codes, the method according to any one of claims 1 to 7 is implemented.

Images