Garbage classification evaluation system

Abstract

The invention discloses a garbage classification evaluation system, which is characterized in that a garbage image corresponding to a target garbage can is acquired, then a garbage classification model and a specific garbage category detection model are used for identifying garbage categories in the garbage image, and whether the identified garbage categories correspond to the target garbage can or not is judged; therefore, the garbage throwing behavior of the user is evaluated. Besides, the garbage classification model and the specific garbage category detection model have different garbage classification identification capabilities, and through cooperation of the garbage classification model and the specific garbage category detection model, the accuracy of a garbage classification evaluation result can be effectively improved.

Description

Technical Field

[0001] The present invention relates to the technical field of garbage classification, and in particular to a garbage classification evaluation system. Background Art

[0002] In recent years, with the rapid growth of population and the continuous development of economy, the amount of domestic waste has shown a continuous increasing trend. Correct waste classification and recycling is of great significance to reducing environmental pollution, improving resource recycling rate, reducing energy waste and building an eco-friendly environment. my country's current waste classification adopts the standard GB / T 19095-2019, which divides domestic waste into four categories: recyclables, kitchen waste, hazardous waste and other waste. However, in practice, due to various factors (such as weak awareness of waste classification among residents, lack of knowledge of waste classification and inability to correctly identify the classification of waste, lack of facilities for waste classification in residential communities, etc.), the actual effect of waste classification is not ideal. In addition, the current waste classification work still relies on manual sorting in waste treatment plants, which is not conducive to the physical and mental health of practitioners, and leads to low efficiency of waste classification, which cannot cope with the growing amount of waste generated in modern society. To this end, the sanitation management department hopes to evaluate the waste classification of residents so that targeted countermeasures can be taken to improve the efficiency of waste classification management. Obviously, it is not realistic to rely on manual evaluation of garbage classification. Therefore, the sanitation management department hopes to use the intelligent means of modern computers to evaluate garbage classification, but there is still a lack of relevant technology in the existing technology. Summary of the invention

[0003] The technical problem to be solved by the present invention is how to realize the evaluation of garbage classification by means of intelligent means of modern computers.

[0004] In order to solve the above technical problems, the first aspect of the present invention discloses a garbage classification evaluation system, the system comprising: An acquisition module, used to acquire a garbage image corresponding to the garbage in a target garbage bin used by a user to throw away garbage at a certain time; A garbage classification recognition module is used to call a garbage classification model to recognize the garbage image and obtain a garbage classification result; A determination module, used for determining whether the waste classification result includes other waste categories different from the target waste category, wherein the target waste category refers to the waste category designated for the target waste bin; An evaluation result recording module, for recording the evaluation result of the user's garbage disposal behavior as unqualified when it is determined that the garbage classification result contains the other garbage categories; A specific garbage category detection module, for, when it is determined that the garbage classification result does not contain the other garbage category, calling a specific garbage category detection model corresponding to the other garbage category to detect whether there is garbage of the other garbage category in the garbage image; The evaluation result recording module is further used to record the evaluation result of the user's garbage throwing behavior as unqualified when it is detected that there is garbage of other garbage categories in the garbage image; The evaluation result recording module is further used to record the evaluation result of the user's garbage throwing behavior as qualified when it is detected that there is no garbage of other garbage categories in the garbage image; A local training module, used for acquiring local training data of the target trash bin, and training the trash classification model based on the local training data; An uploading module, used for uploading the trained local model parameters of the garbage classification model to a global server; An aggregation module, used for the global server to aggregate the local model parameters to obtain global model parameters of the global garbage classification model; The model update module is used to determine whether a preset model update condition is met; when it is determined that the model update condition is met, the global server sends the global model parameters to the target garbage bin locally to update the garbage classification model.

[0005] In the first aspect of the present invention, the specific manner in which the model updating module determines whether the preset model updating condition is met is: Calculating a deviation value between the garbage classification model and the global garbage classification model based on the local model parameters and the global model parameters; When the deviation value is greater than a preset deviation threshold, determining that a model update condition is met; When the deviation value is not greater than the deviation threshold, it is determined that the model update condition is not satisfied.

[0006] In the first aspect of the present invention, the specific manner in which the model updating module calculates the deviation value between the garbage classification model and the global garbage classification model based on the local model parameters and the global model parameters is: in, is the deviation value, is the coefficient, is the number of trash bins, For the The local model parameters uploaded by the trash bin, are global model parameters.

[0007] In the first aspect of the present invention, the specific manner in which the model updating module calculates the deviation value between the garbage classification model and the global garbage classification model based on the local model parameters and the global model parameters is: in, is the deviation value, is the shallow parameter coefficient of the local model, is the deep parameter coefficient of the local model, Less than , is the number of trash bins, For the Shallow parameters of local models uploaded from trash bins, For the The local model deep parameters uploaded by the trash bin, is the shallow parameter of the global model, is the deep parameter of the global model.

[0008] In the first aspect of the present invention, the feature extraction network in the garbage classification model is replaced by a GhostNet module.

[0009] In the first aspect of the present invention, the garbage classification model introduces the CBAM attention mechanism.

[0010] Compared with the prior art, the embodiments of the present invention have the following beneficial effects: In the embodiment of the present invention, a garbage image corresponding to a target garbage bin is obtained, and then the garbage classification model and the specific garbage category detection model are used to identify the garbage category in the garbage image, and by judging whether the identified garbage category corresponds to the target garbage bin, the user's garbage throwing behavior is evaluated. In addition, by making the garbage classification model and the specific garbage category detection model have different garbage classification recognition capabilities, and by the cooperation between the two, the accuracy of the garbage classification evaluation results can be effectively improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0011] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0012] Figure 1 It is a flow chart of a method for evaluating garbage classification disclosed in an embodiment of the present invention; Figure 2 is a schematic diagram of a smart trash can involved in an embodiment of the present invention; Figure 3 is a schematic diagram of model training involved in an embodiment of the present invention; Figure 4 is a schematic diagram of a GhostNet module involved in an embodiment of the present invention; Figure 5 is a schematic diagram of the CBAM attention mechanism involved in an embodiment of the present invention; Figure 6 It is a schematic diagram of the cascade structure of the Ghost convolution and CBAM attention mechanism involved in an embodiment of the present invention; Figure 7 It is a structural diagram of a garbage classification evaluation system disclosed in an embodiment of the present invention; Figure 8 It is a structural diagram of a garbage classification evaluation terminal disclosed in an embodiment of the present invention; Fig. 9 It is a structural schematic diagram of a computer storage medium disclosed in an embodiment of the present invention. DETAILED DESCRIPTION

[0013] In order to enable those skilled in the art to better understand the scheme of the present invention, the technical scheme in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0014] The terms "first", "second", etc. in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish different objects rather than to describe a specific order. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions. For example, a process, method, device, product or end including a series of steps or units is not limited to the listed steps or units, but may optionally include steps or units that are not listed, or may optionally include other steps or units inherent to these processes, methods, products or ends.

[0015] Reference to "embodiments" herein means that a particular feature, structure, or characteristic described in conjunction with the embodiments may be included in at least one embodiment of the present invention. The appearance of the phrase in various places in the specification does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment that is mutually exclusive with other embodiments. It is explicitly and implicitly understood by those skilled in the art that the embodiments described herein may be combined with other embodiments.

[0016] The present invention discloses a method, system, terminal and storage medium for garbage classification evaluation, which obtains a garbage image corresponding to a target garbage bin, and then uses a garbage classification model and a specific garbage category detection model to identify the garbage category in the garbage image, and determines whether the identified garbage category corresponds to the target garbage bin, thereby achieving an evaluation of the user's garbage throwing behavior. In addition, by making the garbage classification model and the specific garbage category detection model have different garbage classification recognition capabilities, and through the cooperation between the two, the accuracy of the garbage classification evaluation results can be effectively improved. The following are detailed descriptions.

[0017] Example 1: Please refer to Figure 1 , Figure 1 FIG. 1 is a flow chart of a method for evaluating garbage classification disclosed in an embodiment of the present invention. Figure 1 As shown, the garbage classification evaluation method may include the following operations: 101. Obtain a garbage image corresponding to garbage in a target garbage bin used by a user to throw away garbage at a certain time.

[0018] In the above step 101, my country divides domestic waste into four categories: recyclables, kitchen waste, hazardous waste and other waste according to the GB / T 19095-2019 standard. In order to achieve waste classification, the sanitation management department usually sets up four types of trash cans in residential communities, which are used to recycle the four major categories of waste. When throwing away garbage, residents can first identify the type of garbage themselves, and then throw different types of garbage into the corresponding types of trash cans. Figure 2 As shown in Figure 1, there are some smart trash cans that can automatically sort garbage. Whether it is done manually by residents or automatically by machines, the sanitation management department hopes to use the intelligent means of modern computers to evaluate the garbage classification. To this end, a camera can be set up in the trash can. Every time garbage enters the trash can, the thrown garbage is photographed, and then the collected garbage image is identified to determine whether the type of garbage is consistent with the type of the trash can, thereby realizing the evaluation of the garbage classification.

[0019] 102. Calling a garbage classification model to identify the garbage image to obtain a garbage classification result.

[0020] In the above step 102, the garbage classification model can use the YOLO model. The YOLO series of algorithms have been widely used in the field of target detection. The YOLO model has a small model size and less computational effort, and has a considerable speed. It is particularly suitable for real-time target detection tasks, and has a high cost-effectiveness without considering whether a GPU exists, especially the CPU occupancy of the PC side is small. The YOLO model is more suitable for the detection of small objects, and the equipment usage cost is low, which is more suitable for the design of the garbage classification evaluation system. The YOLO model mainly consists of four components: input, trunk, neck and head, and can use architectures such as YOLOv5n, YOLOv5s, YOLOv5m, YOLOv5l and YOLOv5x. These architectures differ in network depth and feature map width, allowing users to choose the architecture that best meets their specific needs. In order to meet the actual needs of garbage classification evaluation, such as real-time detection and simple deployment, YOLOv5s with a small model size and fast detection speed can be selected as the basic algorithm. It can provide higher detection accuracy and faster processing speed, providing a reliable foundation for the efficient operation and accurate classification of the intelligent garbage classification and evaluation system.

[0021] 103. Determine whether the garbage classification result includes other garbage categories different from the target garbage category, wherein the target garbage category refers to the garbage category that the target garbage bin is designated to be disposed of.

[0022] In the above step 103, the garbage classification model identifies the garbage image, and after obtaining the garbage classification result, it can be used to determine whether the garbage classification behavior is accurate, thereby realizing the evaluation of garbage classification. For example, when throwing garbage, the garbage is placed in a recyclable garbage bin. After the garbage classification model identifies the garbage image, it is found that the types of garbage are recyclables and kitchen waste. It is determined that there are other types of garbage, and the evaluation result of the user's garbage throwing behavior is recorded as unqualified.

[0023] 104. When it is determined that the garbage classification result includes the other garbage categories, the evaluation result of the user's garbage throwing behavior is recorded as unqualified.

[0024] 105. When it is determined that the garbage classification result does not include the other garbage categories, a specific garbage category detection model corresponding to the other garbage categories is called to detect whether there is garbage of the other garbage categories in the garbage image.

[0025] In the above step 105, due to technical limitations, the classification accuracy of the garbage classification model is usually limited. In order to further improve the accuracy of the garbage classification evaluation, when it is determined that there are no other garbage categories in the garbage classification results, the specific garbage category detection model corresponding to the other garbage categories is called to detect whether there are garbage of other garbage categories in the garbage image. For example, the type of the garbage bin is recyclables, and the classification result of the garbage classification model is also recyclables. At this time, the specific garbage category detection models corresponding to the three categories of kitchen waste, hazardous waste and other garbage can be called to detect the garbage image. If kitchen waste, hazardous waste and other garbage are detected in the garbage image, the evaluation result of the user's garbage throwing behavior is recorded as unqualified. The specific garbage category detection model can also use the YOLO model. However, the training data set of the specific garbage category detection model can be different from that of the garbage classification model, so that the specific garbage category detection model and the garbage classification model have different recognition capabilities. For example, for the specific garbage category detection model for recyclables, its training data set mainly contains images of recyclable garbage, so that the trained specific garbage category detection model can more accurately identify images of recyclable garbage and has stronger recognition capabilities for images of recyclable garbage. The training data set for the garbage classification model contains four types of garbage images, aiming to make the garbage classification model have stronger generalization capabilities and be able to more accurately identify various types of garbage images. The coordination of the garbage classification model and the specific garbage category detection model can improve the accuracy of the garbage classification evaluation results.

[0026] 106. When it is detected that garbage of other garbage categories exists in the garbage image, the evaluation result of the user's garbage throwing behavior is recorded as unqualified.

[0027] 107. When it is detected that there is no garbage of other garbage categories in the garbage image, the evaluation result of the user's garbage throwing behavior is recorded as qualified.

[0028] It can be seen that implementation Figure 1 The described garbage classification evaluation method obtains a garbage image corresponding to a target garbage bin, and then uses a garbage classification model and a specific garbage category detection model to identify the garbage category in the garbage image, and by judging whether the identified garbage category corresponds to the target garbage bin, the user's garbage throwing behavior is evaluated. In addition, by making the garbage classification model and the specific garbage category detection model have different garbage classification recognition capabilities, and by the cooperation between the two, the accuracy of the garbage classification evaluation results can be effectively improved.

[0029] In an optional embodiment, the garbage classification evaluation method further includes: Acquire local training data of the target trash bin, and train the trash classification model based on the local training data; Uploading the trained local model parameters of the garbage classification model to the global server; The global server aggregates the local model parameters to obtain global model parameters of the global garbage classification model; Determine whether a preset model update condition is met; when it is determined that the model update condition is met, the global server sends the global model parameters to the target trash bin locally to update the trash classification model.

[0030] In this optional embodiment, due to the complexity of garbage images, in order to maintain a good recognition accuracy of the garbage classification model and the specific garbage category detection model, more training data is needed to train the model. To this end, the garbage images that are constantly generated in practice can be used as nutrients for the classification model and used for model training. Specifically, Figure 3 As shown, garbage bins distributed in various places can use the garbage images collected locally to train the model. After the training is completed, the local model parameters are uploaded to the global server. Then the global server aggregates the local model parameters to obtain the global model parameters of the global garbage classification model. When the local model needs to be updated, the global server sends the global garbage classification model to garbage bins in various places, thereby updating the local models of garbage bins in various places. In this way, it is possible to avoid transmitting a large amount of training data between the local and global servers, ensuring the efficiency and security of data transmission, and to use the computing power resources of garbage bins distributed in various places to provide computing power resource support for the training of the classification model. At the same time, since the global garbage classification model aggregates the local model parameters of garbage bins distributed in various places, it is equivalent to aggregating the training effects of garbage bins in various places, so that the generalization ability and accuracy of the global garbage classification model recognition are higher, which can avoid overfitting of data on a single device and improve the accuracy and stability of the training model.

[0031] In an optional embodiment, determining whether a preset model update condition is met includes: Calculating a deviation value between the garbage classification model and the global garbage classification model based on the local model parameters and the global model parameters; When the deviation value is greater than a preset deviation threshold, determining that a model update condition is met; When the deviation value is not greater than the deviation threshold, it is determined that the model update condition is not satisfied.

[0032] In this optional embodiment, during the continuous training process, the model parameters of the local model and the global model will inevitably be continuously updated. If the local model and the global model are kept consistent at all times, a large amount of data transmission will inevitably be generated between the local and global models, which is obviously not realistic. Therefore, the conditions for model update can be set. When the conditions for model update are met, the global model will be sent to the local to update the local model. This can avoid a large amount of data transmission and constrain the local model parameters of each device to not be too far away from the global model parameters, which helps to avoid the generation of local optimal solutions and improve the convergence speed and generalization performance of the global model.

[0033] In an optional embodiment, calculating the deviation value between the garbage classification model and the global garbage classification model based on the local model parameters and the global model parameters includes: in, is the deviation value, is the coefficient, is the number of trash bins, For the The local model parameters uploaded by each trash bin, and w are the global model parameters.

[0034] In this optional embodiment, by introducing a regularization term, it is possible to better adapt to unbalanced and non-independent data distribution, prevent over-reliance on the data of certain participants, and enable the global model to better reflect the overall characteristics of the participants.

[0035] In an optional embodiment, calculating the deviation value between the garbage classification model and the global garbage classification model based on the local model parameters and the global model parameters includes: in, is the deviation value, is the shallow parameter coefficient of the local model, is the deep parameter coefficient of the local model, Less than , is the number of trash bins, For the Shallow parameters of local models uploaded from trash bins, For the The local model deep parameters uploaded by the trash bin, is the shallow parameter of the global model, is the deep parameter of the global model.

[0036] In this optional embodiment, during the continuous training process, the shallow parameters of the model are usually updated more frequently than the deep parameters, and the change of the deep parameters has a greater impact on the model recognition ability. Therefore, different weight coefficients can be set for the shallow parameters and deep parameters of the model, so that the weight of the deep parameters is higher than the weight of the shallow parameters, so that the deviation value can more accurately reflect the conditions for model update.

[0037] In an optional embodiment, the preset model update condition may also be the number of training rounds or the amount of training data.

[0038] In this optional embodiment, the number of rounds of model training or the amount of training data can also be used as a condition for model update. Usually, after the model has undergone multiple rounds of training, or the amount of trained data has reached a certain amount, it can be inferred that the model has undergone significant changes, and it is necessary to synchronize the local model and the global model to prevent the local model parameters from being too far away from the global model parameters.

[0039] In an optional embodiment, the feature extraction network in the garbage classification model is replaced by a GhostNet module.

[0040] In this optional embodiment, if Figure 4 As shown in the figure, GhostNet first uses a small number of convolution kernels to extract features from the input feature map, then further performs linear transformation operations on this part of the feature map, and finally generates the final feature map through splicing operations. For the input image, , where c, h, w are the number of input channels, the input image height, and the input image width, respectively. In general, the operation of any convolutional layer that produces n feature maps can be written as in, represents the convolution operation, b represents the bias, , which means the output feature map has n channels, represents the filter of this convolution layer. and is the height and width map of the output features, is the kernel size of the convolution filter f, and the number of FLOPs required in this convolution process can be calculated as , usually hundreds of thousands, because the number of filters n and the number of channels c are usually very large. When using Ghost convolution, the amount of computation is The computational compression ratio is Among them, s is the kernel area of ​​the convolution filter f, and d*d is the kernel size of the second convolution filter f.

[0041] This method reduces the learning cost of non-critical features: that is, by combining a small number of convolution kernels with cheaper linear transformation operations instead of conventional convolution methods, the amount of convolution operations is effectively reduced without affecting the performance of the model.

[0042] In an optional embodiment, the garbage classification model introduces a CBAM attention mechanism.

[0043] In this optional embodiment, CBAM aims to address the limitations of traditional convolutional neural networks in processing information of different scales, shapes, and directions. It introduces two attention mechanisms: channel attention and spatial attention. Channel attention is used to enhance the feature representation of different channels, while spatial attention is used to extract key information at different locations in space. Figure 5 As shown in Figure 2, CBAM consists of two key parts: Channel Attention Module (CAM) and Spatial Attention Module (SAM). These two modules can be embedded in different layers of Convolutional Neural Network (CNN) to enhance feature representation.

[0044] In the channel attention module, the global feature representation of each channel is first calculated through a global average pooling operation. Then, the importance weights of each channel are learned through a fully connected layer (or convolutional layer). Finally, these weights are applied to the feature map of each channel to enhance useful channel information. CAM emphasizes key features, while SAM emphasizes the spatial localization of these key features. The formula of the channel attention module is: in It is a nonlinear The spatial attention module is composed of two linear layers and a ReLU activation function. in, Indicates that the convolution kernel size is 7x7, and Concat represents the concatenation operation. Thus, CBAM can dynamically adjust the importance of each channel and spatial position in the feature map, thereby improving the representation ability and performance of the convolutional neural network.

[0045] like Figure 6As shown in the figure, Ghost convolution and CBAM attention mechanism can be implemented through a cascade structure. The CBAM channel attention mechanism is used to obtain the global average of each channel through global average pooling of the input feature map, and then a set of fully connected layers are used to generate channel attention weights. Then, new weight parameters are generated through the Sigmoid function. These weight parameters are applied to each channel of the input feature map, so that the linear operation makes a greater contribution to the final target feature, thereby obtaining a more effective feature map. The cascade structure of Ghost convolution and CBAM attention mechanism effectively combines the advantages of Ghost convolution and CBAM attention mechanism.

[0046] It can be seen that the garbage classification evaluation method of the embodiment of the present invention obtains the garbage image corresponding to the target garbage bin, and then uses the garbage classification model and the specific garbage category detection model to identify the garbage category in the garbage image, and by judging whether the identified garbage category corresponds to the target garbage bin, the user's garbage throwing behavior is evaluated. In addition, by making the garbage classification model and the specific garbage category detection model have different garbage classification recognition capabilities, and through the cooperation between the two, the accuracy of the garbage classification evaluation results can be effectively improved.

[0047] Example 2: Please refer to Figure 7 , Figure 7 Schematic diagram of a garbage classification evaluation system disclosed in an embodiment of the present invention. Figure 7 As shown, the garbage classification evaluation system may include: The acquisition module 701 is used to acquire the garbage image corresponding to the garbage in the target garbage bin used by the user to throw away the garbage at a certain time; The garbage classification recognition module 702 is used to call the garbage classification model to recognize the garbage image and obtain the garbage classification result; A determination module 703 is used to determine whether the waste classification result includes other waste categories different from the target waste category, wherein the target waste category refers to the waste category designated to be disposed in the target waste bin; The evaluation result recording module 704 is used to record the evaluation result of the user's garbage disposal behavior as unqualified when it is determined that the garbage classification result contains the other garbage categories; The specific garbage category detection module 705 is used to, when it is determined that the garbage classification result does not contain the other garbage category, call the specific garbage category detection model corresponding to the other garbage category to detect whether there is garbage of the other garbage category in the garbage image; The evaluation result recording module 704 is further configured to record the evaluation result of the user's garbage throwing behavior as unqualified when it is detected that garbage of other garbage categories exists in the garbage image; The evaluation result recording module 704 is further configured to record the evaluation result of the user's garbage disposal behavior as qualified when it is detected that there is no garbage of other garbage categories in the garbage image.

[0048] In an optional embodiment, the system further includes: A local training module, used for acquiring local training data of the target trash bin, and training the trash classification model based on the local training data; An uploading module, used for uploading the trained local model parameters of the garbage classification model to a global server; An aggregation module, used for the global server to aggregate the local model parameters to obtain global model parameters of the global garbage classification model; The model update module is used to determine whether a preset model update condition is met; when it is determined that the model update condition is met, the global server sends the global model parameters to the target garbage bin locally to update the garbage classification model.

[0049] In an optional embodiment, the model updating module determines whether the preset model updating condition is met in the following specific manner: Calculating a deviation value between the garbage classification model and the global garbage classification model based on the local model parameters and the global model parameters; When the deviation value is greater than a preset deviation threshold, determining that a model update condition is met; When the deviation value is not greater than the deviation threshold, it is determined that the model update condition is not satisfied.

[0050] In an optional embodiment, the specific manner in which the model updating module calculates the deviation value between the garbage classification model and the global garbage classification model based on the local model parameters and the global model parameters is: in, is the deviation value, is the coefficient, is the number of trash bins, For the The local model parameters uploaded by each trash bin, and w are the global model parameters.

[0051] In an optional embodiment, the specific manner in which the model updating module calculates the deviation value between the garbage classification model and the global garbage classification model based on the local model parameters and the global model parameters is: in, is the deviation value, is the shallow parameter coefficient of the local model, is the deep parameter coefficient of the local model, Less than , is the number of trash bins, For the Shallow parameters of local models uploaded from trash bins, For the The local model deep parameters uploaded by the trash bin, is the shallow parameter of the global model, is the deep parameter of the global model.

[0052] In an optional embodiment, the feature extraction network in the garbage classification model is replaced by a GhostNet module.

[0053] In an optional embodiment, the garbage classification model introduces a CBAM attention mechanism.

[0054] For the specific description of the above-mentioned garbage classification and evaluation system, please refer to the specific description of the above-mentioned garbage classification and evaluation method, which will not be repeated here.

[0055] Example 3: Please refer to Figure 8 , Figure 8 This is a schematic diagram of a garbage classification evaluation terminal disclosed in an embodiment of the present invention. Figure 8 As shown, the garbage classification evaluation terminal may include: A memory 801 storing executable program codes; a processor 802 coupled to the memory 801; The processor 802 calls the executable program code stored in the memory 801 to execute the steps in the garbage classification evaluation method disclosed in the first embodiment of the present invention.

[0056] Embodiment 4: The embodiment of the present invention discloses a computer storage medium, which stores computer instructions. When the computer instructions are called, they are used to execute the steps in the garbage classification evaluation method disclosed in Embodiment 1 of the present invention.

[0057] The device embodiments described above are only illustrative, wherein the modules described as separate components may or may not be physically separated, and the components displayed as modules may or may not be physical modules, i.e., they may be located in one place, or they may be distributed on multiple network modules. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Those of ordinary skill in the art may understand and implement it without creative work.

[0058] Through the specific description of the above embodiments, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a necessary general hardware platform, and of course, can also be implemented by hardware. Based on such an understanding, the above technical solution can be essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, and the storage medium includes a read-only memory (ROM), a random access memory (RAM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), a one-time programmable read-only memory (OTPROM), an electronically erasable rewritable read-only memory (EEPROM), a compact disc (CD-ROM) or other optical disc storage, magnetic disk storage, magnetic tape storage, or any other computer-readable medium that can be used to carry or store data.

[0059] Finally, it should be noted that the garbage classification evaluation method, system, terminal and storage medium disclosed in the embodiments of the present invention are only preferred embodiments of the present invention, which are only used to illustrate the technical solution of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, it should be understood by those skilled in the art that the technical solutions described in the aforementioned embodiments can still be modified, or some of the technical features therein can be replaced by equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A garbage classification evaluation system, characterized in that: The system comprises: An acquisition module, used to acquire a garbage image corresponding to the garbage in a target garbage bin used by a user to throw away garbage at a certain time; A garbage classification recognition module is used to call a garbage classification model to recognize the garbage image and obtain a garbage classification result; A determination module, used for determining whether the waste classification result includes other waste categories different from the target waste category, wherein the target waste category refers to the waste category designated for the target waste bin; An evaluation result recording module, for recording the evaluation result of the user's garbage disposal behavior as unqualified when it is determined that the garbage classification result contains the other garbage categories; A specific garbage category detection module, for, when it is determined that the garbage classification result does not contain the other garbage category, calling a specific garbage category detection model corresponding to the other garbage category to detect whether there is garbage of the other garbage category in the garbage image; The evaluation result recording module is further used to record the evaluation result of the user's garbage throwing behavior as unqualified when it is detected that there is garbage of other garbage categories in the garbage image; The evaluation result recording module is further used to record the evaluation result of the user's garbage throwing behavior as qualified when it is detected that there is no garbage of other garbage categories in the garbage image; A local training module, used for acquiring local training data of the target trash bin, and training the trash classification model based on the local training data; An uploading module, used for uploading the trained local model parameters of the garbage classification model to a global server; An aggregation module, used for the global server to aggregate the local model parameters to obtain global model parameters of the global garbage classification model; The model update module is used to determine whether a preset model update condition is met; when it is determined that the model update condition is met, the global server sends the global model parameters to the target garbage bin locally to update the garbage classification model.

2. The garbage classification evaluation system according to claim 1 is characterized in that: The specific method by which the model updating module determines whether the preset model updating condition is met is: Calculating a deviation value between the garbage classification model and the global garbage classification model based on the local model parameters and the global model parameters; When the deviation value is greater than a preset deviation threshold, determining that a model update condition is met; When the deviation value is not greater than the deviation threshold, it is determined that the model update condition is not satisfied.

3. The garbage classification evaluation system according to claim 2 is characterized in that: The specific manner in which the model updating module calculates the deviation value between the garbage classification model and the global garbage classification model based on the local model parameters and the global model parameters is: in, is the deviation value, is the coefficient, is the number of trash bins, For the The local model parameters uploaded by the trash bin, are global model parameters.

4. The garbage classification evaluation system according to claim 2 is characterized in that: The specific manner in which the model updating module calculates the deviation value between the garbage classification model and the global garbage classification model based on the local model parameters and the global model parameters is: in, is the deviation value, is the shallow parameter coefficient of the local model, is the deep parameter coefficient of the local model, Less than , is the number of trash bins, For the Shallow parameters of local models uploaded from trash bins, For the The local model deep parameters uploaded by the trash bin, is the shallow parameter of the global model, is the deep parameter of the global model.

5. The garbage classification evaluation system according to any one of claims 1 to 4, characterized in that: The feature extraction network in the garbage classification model is replaced by the GhostNet module.

6. The garbage classification evaluation system according to any one of claims 1 to 4, characterized in that: The garbage classification model introduces the CBAM attention mechanism.

Images