Training method, system, device and storage medium for overflowing pot image recognition model

Abstract

The present invention discloses a training method, system, device and storage medium for an overflowing pot image recognition model. The training method for the overflowing pot image recognition model comprises: recognizing an image to be processed; if it is recognized that the image to be processed includes a target pot, using a patch to cover a target area corresponding to the target pot to generate a training data image; constructing a training data set based on a number of the training data images, and using the training data set to train a preset image recognition model to obtain the overflowing pot image recognition model. The training method for the overflowing pot image recognition model provided by the present invention improves the robustness of the training by using a random patch covering method for image enhancement during the training process, so that the overflowing pot image recognition model finally trained can focus on recognizing the bubble state at the edge of the pot, reduce the interference of the pot contents on the overflowing pot state recognition, and improve the model's learning efficiency for the overflowing pot feature area.

Description

Technical Field

[0001] The present invention relates to the field of image processing technology, and in particular to a training method, system, device and storage medium for an overflowing pot image recognition model. Background Art

[0002] During long cooking sessions, users often leave the cooking area midway through. This can result in users being unable to detect and address overflowing situations in a timely manner, potentially leading to accidents such as gas leaks and posing serious safety risks. Therefore, how to better identify overflowing during cooking has become a focus of current research.

[0003] With the improvement of computer equipment performance and technological advancements, deep learning-based image processing has become the mainstream solution for pot overflow detection. Currently, cameras are usually used to capture real-time images of the pot's cooking status, and image recognition algorithms are used to determine whether the pot is overflowing.

[0004] However, the image recognition algorithm for judging the overflowing state in the prior art recognizes the entire collected image, which consumes a large amount of computing resources during the processing, is inefficient, and has low recognition accuracy. Summary of the Invention

[0005] The technical problem to be solved by the present invention is to overcome the defects in the existing technology of image recognition algorithms for judging the overflowing pot status that require a large amount of computing resources during the processing process, have low efficiency and low recognition accuracy, and provide a training method, system, equipment and storage medium for an overflowing pot image recognition model.

[0006] The present invention solves the above technical problems through the following technical solutions:

[0007] In a first aspect, the present invention provides a method for training an overflowing pot image recognition model, the method comprising:

[0008] Recognize the image to be processed;

[0009] If it is recognized that the image to be processed includes a target cookware, covering a target area corresponding to the target cookware with a patch to generate a training data image;

[0010] A training data set is constructed based on a number of the training data images, and a preset image recognition model is trained using the training data set to obtain the overflowing pot image recognition model.

[0011] Preferably, the step of covering the target area corresponding to the target cookware with a patch specifically includes:

[0012] Identify the location and type of the target cookware using a preset target detection algorithm and mark it to determine the target area;

[0013] The center point coordinates, width and height of the patch are randomly determined to patch the target area. The center point coordinates (x c ,y c )Width w t and height h t The determination is as follows:

[0014]

[0015]

[0016]

[0017]

[0018] Wherein, w represents the width of the image to be processed, h represents the height of the image to be processed, and random() represents selecting a random number from a preset range.

[0019] Preferably, the step of constructing a training data set based on the plurality of training data images specifically includes:

[0020] performing image enhancement processing and / or multi-scale processing on the training data image;

[0021] constructing the training data set based on the processed training data images;

[0022] The image enhancement processing includes at least one of HSV color space enhancement, random upside down flipping and random cropping.

[0023] Preferably, before the step of performing enhancement processing on the training data image, the method further includes:

[0024] Generate random numbers;

[0025] If the random number is greater than a preset threshold, the step of performing image enhancement processing and / or multi-scale processing on the training data image is executed.

[0026] Preferably, the training method of the overflowing pot image recognition model further includes:

[0027] Sampling training video data including the target cookware to use the sampled image frames as the images to be processed;

[0028] If it is determined according to the sampled image frames that the target cookware is in an overflowing state, the training video data is uniformly sampled; otherwise, the training video data is non-uniformly sampled.

[0029] Preferably, the calculation formula for the sampling frequency of the non-uniform sampling is:

[0030]

[0031] Wherein, θ represents the probability of each frame of the training video data being selected in the non-overflowing state, x i represents the current image frame, t represents the probability factor, x s represents the starting image frame corresponding to the non-overflowing state in the training video data, x e Indicates the ending image frame corresponding to the non-overflowing state in the training video data.

[0032] In a second aspect, the present invention provides a method for identifying an overflowing pot image, the method comprising:

[0033] Acquire an image to be identified including a target pot;

[0034] The image data to be recognized is input into the overflowing pot image recognition model trained by the training method of the overflowing pot image recognition model as described above, so as to recognize whether the target pot is in an overflowing pot state.

[0035] In a third aspect, the present invention provides a range hood, wherein a camera module is provided at the central axis of the range hood facing the stove, and the angle between the camera module and the horizontal plane is less than or equal to 10°, and the camera module is used to collect images to be processed including the target pots on the stove.

[0036] In a fourth aspect, the present invention provides an electronic device comprising a memory, a processor, and a computer program stored in the memory and runnable on the processor, wherein when the processor executes the computer program, the training method for the overflowing pot image recognition model as described above, or the recognition method for the overflowing pot image as described above is implemented.

[0037] In a fifth aspect, the present invention provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the training method of the overflowing pot image recognition model as described above, or the method for recognizing the overflowing pot image as described above.

[0038] The positive progressive effect of the present invention is that: the training method of the overflowing pot image recognition model provided by the present invention improves the robustness of training by using a random patch covering method to perform image enhancement during the training process, so that the overflowing pot image recognition model finally trained can focus on identifying the bubble state at the edge of the pot, thereby reducing the interference of the pot contents on the overflowing pot state recognition, improving the model's learning efficiency for the overflowing pot feature area, and being able to obtain a model with a higher accuracy in identifying the pot moving state, saving computing resources, and achieving efficient and high-accuracy recognition. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] Figure 1 Schematic diagram of the flow of the training method of the overflowing pot image recognition model in Example 1 of the present invention.

[0040] Figure 2 Schematic diagram of the specific process of step S1 of the training method of the overflowing pot image recognition model in Example 1 of the present invention.

[0041] Figure 3 This is a flow chart of the first part of the training method for the overflowing pot image recognition model of Example 1 of the present invention.

[0042] Figure 4 This is a flow chart of the second part of the training method for the overflowing pot image recognition model of Example 1 of the present invention.

[0043] Figure 5 Schematic diagram of the flow of the overflowing pot image recognition method of Example 2 of the present invention.

[0044] Figure 6 This is a structural diagram of an electronic device for implementing the training method of the overflowing pot image recognition model of Example 1 or the overflowing pot image recognition method of Example 2 according to Example 4 of the present invention. DETAILED DESCRIPTION

[0045] The present invention is further described below by way of examples, but the present invention is not limited to the scope of the examples.

[0046] Example 1

[0047] In cooking scenarios, the occurrence of overflowing pots is often closely related to the type of ingredients being cooked and the level of heat. However, when overflowing, a large amount of water vapor and foam usually forms at the edge of the pot, and the ingredients are often submerged in the bubbles. Therefore, it is difficult to directly determine whether the pot is overflowing based on the type of ingredients and the level of heat. In particular, when the pot has a lid, especially an opaque lid, the judgment of whether the pot is overflowing often relies heavily on the judgment of the bubbles at the edge of the pot.

[0048] Therefore, in order to reduce the interference of the color and shape characteristics of the pot contents and the pot lid itself on the overflow pot recognition itself, this embodiment discloses a training method for an overflow pot image recognition model. By randomly patching the contents in the pot in the training data before model training, it is ensured that the influence of the image characteristics of the contents on the overflow pot recognition is reduced, thereby improving the model's learning of the overflow pot feature area.

[0049] like Figure 1 As shown, the training method of the overflowing pot image recognition model includes:

[0050] S1, identify the image to be processed;

[0051] S2. If it is recognized that the image to be processed includes a target cookware, a target area corresponding to the target cookware is covered with a patch to generate a training data image;

[0052] S3. Construct a training data set based on a number of training data images, and use the training data set to train a preset image recognition model to obtain an overflowing pot image recognition model.

[0053] In step S1, this embodiment utilizes target detection and classification results. First, target detection is performed to determine whether the target pan is included in the image to be processed. Specifically, existing target detection algorithms in deep learning can be used for this target detection, and the type of target detection algorithm is not specifically limited herein.

[0054] In a preferred embodiment, Figure 2 As shown, step S1 specifically includes:

[0055] S11. Identify the location and type of the target pot using a preset target detection algorithm and mark it to determine the target area.

[0056] S12, randomly determine the center point coordinates, width and height of the patch to patch the target area, the center point coordinates (x c ,y c )Width w t and height h t The determination is as follows:

[0057]

[0058]

[0059]

[0060]

[0061] Where w represents the width of the image to be processed, h represents the height of the image to be processed, and random() represents selecting a random number from a preset range.

[0062] In this example, cookware is categorized into six common types: woks, frying pans, milk pans, boiling pots, pressure cookers, and casseroles. By collecting images of different types of cookware and annotating the locations and labels of the target cookware using common annotation software, and then using common object detection algorithms to detect the target cookware's location, the target cookware's shape and other characteristics can be used to more accurately determine the area of the image to be processed where the target cookware resides, improving recognition efficiency and accuracy.

[0063] In a specific implementation, a camera module positioned on the central axis of the range hood facing the stovetop can be used to capture an initial image of a target pot on the stovetop. Labeling software can then be used to create a pot dataset to generate training data images containing the pot's type and location coordinates. This allows the location and type of the target pot to be identified based on the training data images. The angle between the camera module and the horizontal plane is less than or equal to 10°.

[0064] Regarding step S2, in the specific implementation process, by determining the coordinate position of the center point of the patch (x c ,y c ) and its specific width w t and height h t , the specific setting position of the patch can be uniquely determined, so that the above-mentioned training data image can be generated.

[0065] In step S3, the training image data obtained in step S2 are input into an existing image recognition model (InceptionV3 is used in this embodiment) for model training to obtain an overflowing pot image recognition model. It should be noted that the classification model selected here is only used for illustration and is not limited thereto.

[0066] In a preferred embodiment, before step S3, as Figure 3 As shown, it also includes:

[0067] S31, performing image enhancement processing and / or multi-scale processing on the training data image;

[0068] S32, constructing a training data set based on the processed training data image;

[0069] The image enhancement processing includes at least one of HSV color space enhancement, random upside-down flipping, and random cropping.

[0070] In this embodiment, the image to be processed is collected by a camera module installed on the range hood. During the installation of the range hood, there is usually an installation error range of about 5 cm between the hook and the stove surface. This may lead to different detection accuracy when using different devices and different models. For this reason, in this embodiment, an enhanced strategy is used to reduce the impact of different equipment installation positions on the detection accuracy.

[0071] Specifically, in order to enhance the generalization ability of the model algorithm finally obtained, this embodiment can perform image enhancement processing on the training data image when constructing the training data set, that is, to imitate the changes in image hue, contrast and position caused by changes in natural ambient light and artificial light in real scenes, at least one of the common enhancement processing such as HSV color space enhancement, random upside down flipping, random cropping, etc.

[0072] For multi-scale processing, let the training data image size be s, set the multi-scale scaling factor σ, and set the maximum step size in the preset image recognition model to stride, then

[0073]

[0074] Here, random represents a random value function, that is, 0.5-1.5 times the size of the training data image is selected as the size of the scaled image, and the specific value is randomly selected; the maximum step size stride is set so that when the size of the input training data image changes, the size type of the enhanced image output by the network after the convolution operation with the maximum step size is integer.

[0075] This embodiment takes the image enhancement and multi-scale processing of the training data image as an example for detailed description. Therefore, the processed training data image n is finally obtained. s It can be expressed as:

[0076] n s =s*σ

[0077] By processing the training image data n s Constructing it as a training data set can effectively expand the amount of training data and the scope of scenarios in the training data set, thereby improving the recognition accuracy of the overflowing pot image recognition model finally trained.

[0078] Preferably, before step S31, Figure 4 As shown, it also includes:

[0079] S301, generate a random number;

[0080] S302: If the random number is greater than the preset threshold, execute step S31.

[0081] Before performing enhancement processing, a random number threshold T is first set to decide whether to perform image enhancement processing and / or multi-scale processing. That is, when the generated random number r>T, the processed training data image is used for training, otherwise the original image is used for training.

[0082] In this embodiment, through image enhancement processing and multi-scale processing combined with random cropping, the different target sizes caused by the different installation heights of the range hood can be simulated, thereby significantly improving the detection effect under different height fields of view and improving the recognition accuracy.

[0083] In a preferred embodiment, the training method of the overflowing pot image recognition model further includes:

[0084] Sampling training video data including the target cookware to use the sampled image frames as the images to be processed;

[0085] If it is determined according to the sampled image frames that the target cookware is in an overflowing state, the training video data is uniformly sampled; otherwise, the training video data is non-uniformly sampled.

[0086] Before importing the detected target pot area into the trained overflowing pot image recognition algorithm for overflowing pot recognition, the training method of the overflowing pot image recognition model in this embodiment further includes generating an overflowing pot recognition training data set for calculating the image recognition model.

[0087] Specifically, the training data for pot overflow recognition mainly corresponds to two categories: pot overflow state and pot non-overflow state. Since the duration of pot overflow in real-world scenarios is much shorter than the normal cooking time, the amount of training data corresponding to the pot overflow state is often greater. To balance the two, in actual data sampling, the sampling frequency in the pot overflow state is related to the sampling frequency in the normal cooking state and the pot overflow duration is shorter. Therefore, in the actual sampling process, the pot overflow stage adopts uniform sampling. In the early stage of normal cooking, the cooking image features are significantly different from the image features of the pot overflow, and the cooking image features before the pot overflow are more difficult to distinguish from the image features of the pot overflow. Therefore, a non-uniform sampling strategy of "sparse at the beginning and dense at the end" is adopted in the normal cooking range.

[0088] In this embodiment, the overflow state adopts the alternate frame sampling strategy, and the sampling frequency is set to h. Then the number of samples is:

[0089]

[0090] Among them, n Indicates the number of samples obtained during the overflow phase, o s Indicates the starting image frame of the training video data corresponding to the overflowing state, o e Indicates the end image frame of the training video data corresponding to the overflowing pot state.

[0091] The calculation formula for the sampling frequency of non-uniform sampling in the non-overflow stage is:

[0092]

[0093] Among them, θ represents the probability of each frame being selected in the training video data under the non-overflowing state, x i represents the current image frame, t represents the probability factor, x s represents the starting image frame corresponding to the non-overflowing state in the training video data, x e Indicates the ending image frame corresponding to the non-overflowing state in the training video data.

[0094] Based on the distinction between the above-mentioned different sampling methods, it can effectively ensure that the number of samples in the normal cooking stage is similar to the number of samples in the overflowing stage, and the recognized time can be close to the start time of the overflowing state, thereby improving the recognition efficiency of the model finally obtained by training.

[0095] The training method of the overflowing pot image recognition model provided in this embodiment improves the robustness of training by using a random patch covering method to perform image enhancement during the training process, so that the overflowing pot image recognition model finally trained can focus on identifying the bubble state at the edge of the pot, thereby reducing the interference of the pot contents on the overflowing pot state identification, improving the model's learning efficiency for the overflowing pot feature area, and being able to obtain a model with a higher accuracy in identifying the pot moving state, saving computing resources, and achieving efficient and high-accuracy recognition.

[0096] Example 2

[0097] This embodiment discloses a method, such as Figure 5 As shown, the method for identifying the overflowing pot image includes:

[0098] S21, obtaining an image to be identified including a target cookware;

[0099] S22. Input the image data to be recognized into the overflowing pot image recognition model trained by the training method of the overflowing pot image recognition model in Example 1 to recognize whether the target pot is in an overflowing state.

[0100] Specifically, this embodiment also uses a range hood including a camera module as described in Example 1 to collect images of the cooking status of the target pot on the stove in real time. Based on the overflowing pot image recognition model obtained in Example 1, when the camera module captures that the target pot is currently in an overflowing state, the buzzer of the range hood itself is used to sound an alarm, and at the same time, possible methods such as App reminders are used to remind the user that the overflowing pot may occur, and the user is advised to deal with it in time.

[0101] The overflowing pot image recognition method provided in this embodiment performs overflowing pot image recognition by using the overflowing pot image recognition model trained by the training method of the overflowing pot image recognition model of Example 1, so that it can focus on identifying the bubble state at the edge of the pot, thereby reducing the interference of the pot contents on the overflowing pot state recognition, improving the recognition efficiency of the overflowing pot characteristic area, saving computing resources, and achieving efficient and high-accuracy recognition.

[0102] Example 3

[0103] This embodiment discloses a range hood, which has a camera module provided at the central axis of the range hood facing the stove. The angle between the camera module and the horizontal plane is less than or equal to 10°. The camera module is used to collect images to be processed including target pots on the stove.

[0104] In recent years, smart home appliances have gradually come into people's view. However, due to the relatively closed environment of the kitchen appliance industry, it is difficult to connect with external data. Due to this limitation, the level of intelligence of kitchen appliances is relatively weak.

[0105] The range hood of this embodiment uses a camera installed on the central axis facing the stovetop to monitor the cooking status of the cooktop area. The collected images are then processed using the overflowing pot image recognition model trained in Example 1 to provide feedback to the user. In this way, the user can monitor the current cooking status in real time, reducing potential risks.

[0106] Specifically, the camera is installed on the central axis of the range hood to ensure that the left and right stove areas are on both sides of the camera's image acquisition field of view. In addition, to ensure that the camera's imaging angle can completely cover the entire stove area, when the pot is cooking normally, the camera's field of view (FOV) can cover the entire upper surface of the pot. Therefore, the inclination angle of the camera module to the horizontal plane is less than or equal to 10°, and the size of the camera lens and the size and width of the entire module do not exceed the space that can be accommodated by the range hood's own structural parts. To ensure imaging quality, the camera's pixel size is above 100W pixels (set to 100W in this embodiment, but it is not limited to this), the peak signal-to-noise ratio is not less than 35db, and the distortion is less than or equal to 15%.

[0107] Considering that overflowing pot recognition requires real-time processing of images of the current cooking scene, a high-performance computing unit is required for processing. Continuously and stably performing model inference by the computing unit requires ensuring that the chip unit has a good external heat dissipation environment. Therefore, in this embodiment, the development board is integrated on the power board of the range hood, and its CPU and NPU (data processing unit) are connected to the heat sink of the power board through a heat dissipation copper tube to accelerate heat dissipation and ensure recognition stability.

[0108] The range hood in this embodiment can collect image data for the training method of the overflowing pot image recognition model of Example 1, or for the overflowing pot image recognition method of Example 2 to perform overflowing pot image recognition, and has good heat dissipation characteristics, thereby providing effective support for achieving efficient and high-accuracy overflowing pot image recognition.

[0109] Example 4

[0110] Figure 6 This is a schematic diagram of the structure of an electronic device provided in Example 4 of the present invention. The electronic device includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, it implements the method for training the overflowing pot image recognition model provided in Example 1 or the method for recognizing overflowing pot images provided in Example 2. Figure 6 The electronic device 40 shown is only an example and should not limit the functionality and scope of use of the embodiments of the present invention.

[0111] like Figure 6 As shown, the electronic device 40 may be a general-purpose computing device, such as a server device. Components of the electronic device 40 may include, but are not limited to, the at least one processor 41, the at least one memory 42, and a bus 43 connecting different system components (including the memory 42 and the processor 41).

[0112] The bus 43 includes a data bus, an address bus, and a control bus.

[0113] The memory 42 may include a volatile memory, such as a random access memory (RAM) 421 and / or a cache memory 422 , and may further include a read-only memory (ROM) 423 .

[0114] The memory 42 may also include a program / utility 425 having a set (at least one) of program modules 424, such program modules 424 including but not limited to: an operating system, one or more application programs, other program modules, and program data, each of which or some combination may include an implementation of a network environment.

[0115] The processor 41 executes various functional applications and data processing by running the computer program stored in the memory 42, such as the training method of the overflowing pot image recognition model provided in Example 1 of the present invention, or the overflowing pot image recognition method provided in Example 2.

[0116] The electronic device 40 can also communicate with one or more external devices 44 (e.g., a keyboard, pointing device, etc.). This communication can occur via an input / output (I / O) interface 45. Furthermore, the model-generating device 40 can also communicate with one or more networks (e.g., a local area network (LAN), a wide area network (WAN), and / or a public network, such as the Internet) via a network adapter 46. As shown, the network adapter 46 communicates with other modules of the model-generating device 40 via a bus 43. It should be understood that, although not shown, other hardware and / or software modules can be used in conjunction with the model-generating device 40, including but not limited to microcode, device drivers, redundant processors, external disk drive arrays, RAID (RAID) systems, tape drives, and data backup storage systems.

[0117] It should be noted that although several units / modules or sub-units / modules of the electronic device are mentioned in the detailed description above, this division is merely exemplary and not mandatory. In fact, according to embodiments of the present invention, the features and functions of two or more units / modules described above may be embodied in a single unit / module. Conversely, the features and functions of a single unit / module described above may be further divided and embodied by multiple units / modules.

[0118] Example 5

[0119] This embodiment provides a computer-readable storage medium having a computer program stored thereon. When the program is executed by a processor, the training method of the overflowing pot image recognition model provided in Example 1 or the method for recognizing the overflowing pot image provided in Example 2 is implemented.

[0120] The readable storage medium may include, but is not limited to, a portable disk, a hard disk, a random access memory, a read-only memory, an erasable programmable read-only memory, an optical storage device, a magnetic storage device, or any suitable combination thereof.

[0121] In a possible embodiment, the present invention can also be implemented in the form of a program product, which includes a program code. When the program product is run on a terminal device, the program code is used to enable the terminal device to execute the training method of the overflowing pot image recognition model provided in Example 1, or the overflowing pot image recognition method provided in Example 2.

[0122] The program code for executing the present invention may be written in any combination of one or more programming languages, and may be executed entirely on the user device, partially on the user device, as an independent software package, partially on the user device and partially on a remote device, or entirely on the remote device.

[0123] Although specific embodiments of the present invention have been described above, those skilled in the art will appreciate that these are merely illustrative and that the scope of the present invention is defined by the appended claims. Those skilled in the art may make various changes or modifications to these embodiments without departing from the principles and essence of the present invention, and such changes and modifications are intended to fall within the scope of the present invention.

Claims

1. A training method for an overflowing pot image recognition model, characterized in that: The training method of the overflowing pot image recognition model includes: Recognize the image to be processed; If it is recognized that the image to be processed includes a target cookware, covering a target area corresponding to the target cookware with a patch to generate a training data image; constructing a training data set based on a number of the training data images, and using the training data set to train a preset image recognition model to obtain the overflowing pot image recognition model; The step of using a patch to cover the target area corresponding to the target cookware specifically includes: Identify the location and type of the target cookware using a preset target detection algorithm and mark it to determine the target area; The center point coordinates, width and height of the patch are randomly determined to patch the target area. The center point coordinates ( )width and height The determination of is as follows: in, Indicates the width of the image to be processed, Indicates the height of the image to be processed, and random() indicates selecting a random number from a preset range; A camera module is provided at the central axis of the range hood facing the stove, and the angle between the camera module and the horizontal plane is less than or equal to 10°. The camera module is used to collect images to be processed including the target pot on the stove.

2. The method for training an overflowing pot image recognition model according to claim 1, wherein: The step of constructing a training data set based on the plurality of training data images specifically includes: performing image enhancement processing and / or multi-scale processing on the training data image; constructing the training data set based on the processed training data images; The image enhancement processing includes at least one of HSV color space enhancement, random upside down flipping and random cropping.

3. The method for training an overflowing pot image recognition model according to claim 2, wherein: Before the step of performing enhancement processing on the training data image, the method further includes: Generate random numbers; If the random number is greater than a preset threshold, the step of performing image enhancement processing and / or multi-scale processing on the training data image is executed.

4. The method for training an overflowing pot image recognition model according to claim 1, wherein: The training method of the overflowing pot image recognition model further includes: Sampling training video data including the target cookware to use the sampled image frames as the images to be processed; If it is determined according to the sampled image frames that the target cookware is in an overflowing state, the training video data is uniformly sampled; otherwise, the training video data is non-uniformly sampled.

5. The method for training an overflowing pot image recognition model according to claim 4, wherein: The calculation formula of the sampling frequency of the non-uniform sampling is: , in, represents the probability of each image frame being selected in the training video data in the non-overflowing state, Indicates the current image frame, represents the probability factor, represents the starting image frame corresponding to the non-overflowing state in the training video data, Indicates the ending image frame corresponding to the non-overflowing state in the training video data.

6. A method for identifying overflowing pot images, characterized in that: The method for identifying the overflowing pot image includes: Acquire an image to be identified including a target cookware; The image data to be recognized is input into an overflowing pot image recognition model trained by the training method of the overflowing pot image recognition model according to any one of claims 1 to 5, so as to identify whether the target pot is in an overflowing pot state.

7. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the computer program, it implements the training method of the overflowing pot image recognition model as described in any one of claims 1 to 5, or the recognition method of the overflowing pot image as described in claim 6.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, it implements the training method of the overflowing pot image recognition model as described in any one of claims 1 to 5, or the recognition method of the overflowing pot image as described in claim 6.

Images