Depression identification method based on house tree person drawing and deep learning, computer device and storage medium

Abstract

The invention discloses a depression identification method based on house tree person drawing and deep learning, a computer device and a storage medium, and the method comprises the steps: inputting house tree person drawing information into a drawing-based feature encoder for processing, and obtaining multiple pieces of group-based feature information, and inputting the group base feature information into a guidance base feature gating unit for processing to obtain group gate information and a base vector, and inputting the group gate information and the base vector into an adaptive channel feature fusion network for processing to obtain a depression identification result. According to the depression early-stage self-recognition tool based on artificial intelligence deep learning, early-stage recognition of depression is achieved under the conditions that interaction between people is not needed and strict adherence is achieved, and the risk that privacy information of recognized people is leaked can be reduced; the method is loose in application condition and easy to popularize and apply to identified people of different age groups. The method is widely applied to the technical field of image processing.

Description

Technical Field

[0001] The present invention relates to the field of image processing technology, and in particular to a depression recognition method based on house-tree-person drawing and deep learning, a computer device and a storage medium. Background Art

[0002] Depression is a mental illness. Identifying depression is helpful for treatment and medical research. With the development of artificial intelligence technology, some related technologies currently apply artificial intelligence technology to depression identification. For example, artificial intelligence technology is used to identify various expressions imitated by subjects, or to induce emotions in subjects through videos, detect dynamic expression changes in subjects or characteristics of electrocardiogram (ECG), electrodermal activity (EDA) and respiratory activity (RSP) physiological signals, so as to determine whether the subject has depression; or use artificial intelligence technology to identify brain maps (such as magnetic resonance imaging, MRI, etc.), and make diagnoses based on brain maps; there are also some related technologies that use artificial intelligence technology to detect social media content of subjects, so as to determine whether the subject has depression.

[0003] The above-mentioned related technologies generally have shortcomings such as many restrictions on use, poor flexibility, and easy to generate other problems. For example, the recognition of imitation expressions and dynamic expression changes induced by videos requires the subjects to strictly abide by specific collection conditions, such as the subject's face facing the camera, but most people watch videos in a comfortable way, and specific conditions may cause discomfort to the subjects, thereby affecting the quality of the data; the collection of physiological signals and the recognition of brain maps require special equipment, which is not suitable for self-testing by the general population and cannot be widely promoted; using social media content of people with depression to design deep learning methods requires a certain degree of privacy infringement, and data leakage will cause major social problems. Summary of the invention

[0004] In view of the technical problems of current depression recognition technology based on artificial intelligence, such as many usage restrictions, poor flexibility, and easy to derive other problems, the purpose of the present invention is to provide a depression recognition method, computer device and storage medium based on house-tree-person drawing and deep learning.

[0005] On the one hand, an embodiment of the present invention includes a depression identification method based on house-tree-person drawing and deep learning, and the depression identification method based on house-tree-person drawing and deep learning includes the following steps:

[0006] Get house-tree-person drawing information;

[0007] Obtaining a painting-based feature encoder, a directional-based feature gating unit, and an adaptive channel feature fusion network;

[0008] Inputting the house-tree-person painting information into the painting-based feature encoder for processing, and obtaining a plurality of group-based feature information output by the painting-based feature encoder;

[0009] Input each of the group basis feature information into the directional basis feature gating unit for processing, and obtain the group gate information and multiple basis vectors output by the directional basis feature gating unit; wherein each of the basis vectors corresponds to a corresponding one of the group basis feature information;

[0010] The group gate information and each of the basis vectors are input into the adaptive channel feature fusion network for processing, and a depression recognition result output by the adaptive channel feature fusion network is obtained.

[0011] Furthermore, the method of obtaining the painting base feature encoder, the directional base feature gating unit and the adaptive channel feature fusion network includes:

[0012] Establishing a first convolutional layer, at least two residual modules and at least two grouped residual modules; wherein the residual module includes a plurality of second convolutional layers, and the grouped residual module includes a plurality of grouped convolutional layers;

[0013] The first convolutional layer, each of the residual modules and each of the grouped residual modules are connected in sequence to obtain the painting-based feature encoder.

[0014] Furthermore, the method of obtaining the painting base feature encoder, the directional base feature gating unit and the adaptive channel feature fusion network includes:

[0015] Create global average pooling layers, linear layers, and activation layers;

[0016] The global average pooling layer, the linear layer and the activation layer are connected in sequence to obtain the directional base feature gating unit.

[0017] Furthermore, the method of obtaining the painting base feature encoder, the directional base feature gating unit and the adaptive channel feature fusion network includes:

[0018] A graph attention network is established as the adaptive channel feature fusion network.

[0019] Further, the inputting each of the group basis feature information into the directional basis feature gating unit for processing, and obtaining the group gate information and multiple basis vectors output by the directional basis feature gating unit, comprises:

[0020] Using the global average pooling layer to process each of the group basis feature information, to obtain the basis vector corresponding to each of the group basis feature information;

[0021] Using the linear layer to process each of the basis vectors, to obtain a query vector corresponding to each of the basis vectors;

[0022] Obtaining the key vectors corresponding to the basis vectors;

[0023] The query vectors are used to form a query matrix, and the key vectors are used to form a key matrix;

[0024] Determining a group attention vector based on the query matrix and the key matrix;

[0025] Processing the group of attention vectors using the activation layer;

[0026] The output result of the activation layer is processed according to a preset threshold to obtain the group gate information.

[0027] Furthermore, the processing of the output result of the activation layer according to a preset threshold to obtain the group gate information includes:

[0028] Traversing each element in the output result of the activation layer;

[0029] For any traversed element, compare the value of the element with the preset threshold;

[0030] When the value of the element is greater than the preset threshold, the value of the element is assigned to 1, otherwise, the value of the element is assigned to 0.

[0031] Furthermore, the inputting the group gate information and each of the basis vectors into the adaptive channel feature fusion network for processing to obtain a depression recognition result output by the adaptive channel feature fusion network includes:

[0032] Determine an adjacency matrix according to the group gate information and each of the basis vectors;

[0033] The graph attention network is used to perform inter-group feature fusion processing on the adjacency matrix to obtain the depression recognition result.

[0034] Further, determining the adjacency matrix according to the group gate information and each of the basis vectors includes:

[0035] Filtering each of the basis vectors according to the group gate information to determine the filtered basis vectors and the retained basis vectors;

[0036] The retained basis vectors are connected to obtain the adjacency matrix.

[0037] On the other hand, an embodiment of the present invention also includes a computer device, including a memory and a processor, the memory is used to store at least one program, and the processor is used to load at least one program to execute the depression identification method based on house-tree-person drawing and deep learning in the embodiment.

[0038] On the other hand, an embodiment of the present invention also includes a computer-readable storage medium, which stores a program executable by a processor. When the program executable by the processor is executed by the processor, it is used to execute the depression identification method based on house-tree-person drawing and deep learning in the embodiment.

[0039] The beneficial effects of the present invention are as follows: the depression identification method based on house-tree-person drawing and deep learning in the embodiment provides an early self-identification tool for depression based on artificial intelligence deep learning, which can make an objective assessment of one's own mental state and achieve early identification of depression by identifying house-tree-person drawing information without the need for interpersonal interaction and compliance with strict conditions; since the house-tree-person drawing information is freely created by the identified person, the risk of leakage of the identified person's privacy information can be reduced; the application conditions of the depression identification method based on house-tree-person drawing and deep learning in the embodiment are loose, so it is easy to promote and apply to identified persons of different age groups. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] Figure 1 Schematic diagram of the steps of a method for identifying depression based on house-tree-person drawing and deep learning in an embodiment;

[0041] Figure 2 It is a schematic diagram of the principle of a depression identification method based on house-tree-person drawing and deep learning in an embodiment;

[0042] Figure 3 Schematic diagram of the structure and principle of the painting-based feature encoder in the embodiment;

[0043] Figure 4 Schematic diagram of the structure and principle of the directional base feature gating unit in the embodiment;

[0044] Figure 5 Schematic diagram of the structure and principle of the adaptive channel feature fusion network in the embodiment. DETAILED DESCRIPTION

[0045] In this embodiment, refer to Figure 1 ,The depression recognition method based on house-tree-person drawing and deep learning includes the following steps:

[0046] S1. Get house-tree-person drawing information;

[0047] S2. Obtain a painting-based feature encoder, a directional-based feature gating unit, and an adaptive channel feature fusion network;

[0048] S3. The house-tree-person painting information is input to the painting-based feature encoder for processing, and the painting-based feature encoder outputs a plurality of groups of base feature information;

[0049] S4. Input each group of base feature information into the directional base feature gating unit for processing, and obtain the group gate information and multiple basis vectors output by the directional base feature gating unit; wherein each basis vector corresponds to a corresponding group of base feature information;

[0050] S5. Input the group gate information and each basis vector into the adaptive channel feature fusion network for processing, and obtain the depression recognition result output by the adaptive channel feature fusion network.

[0051] In this embodiment, steps S1-S5 can all be executed using a computer, that is, all steps in the depression identification method based on house-tree-person drawing and deep learning can be implemented by a computer or other device.

[0052] In step S1, the obtained home, tree, person drawing information (Home, Tree, Person Drawing) may be a drawing drawn by the identified person including images of houses, trees, and people, etc. Specifically, the identified person may draw on paper and then obtain the home, tree, and person drawing information by scanning the drawing paper, or the identified person may draw on electronic drawing paper to collect the drawing action and directly generate the home, tree, and person drawing information.

[0053] Specifically, when executing step S1, multiple house-tree-person painting information can be obtained. The content and size of each house-tree-person painting information are presented by the painter (identified person) according to his or her own emotions. Therefore, step S1 does not use cropping, scaling, and other data enhancement methods that may destroy the painting itself for preprocessing. In order to enhance the data without destroying the content in the painting, when executing step S1, only horizontal flipping, vertical flipping, and horizontal-vertical flipping can be used to preprocess the house-tree-person painting information. Before executing steps S2-S5, all house-tree-person painting information can be unified into the same size (for example, 300x200), and all house-tree-person painting information can be normalized.

[0054] In this embodiment, when executing step S2, that is, obtaining the painting base feature encoder, the directional base feature gating unit and the adaptive channel feature fusion network, the following steps can be specifically performed:

[0055] S201. Establishing a first convolutional layer, at least two residual modules and at least two grouped residual modules; wherein the residual module includes a plurality of second convolutional layers, and the grouped residual module includes a plurality of grouped convolutional layers;

[0056] S202. Connect the first convolutional layer, each residual module and each grouped residual module in sequence to obtain a painting-based feature encoder;

[0057] S203. Establish a global average pooling layer, a linear layer and an activation layer;

[0058] S204. Connect the global average pooling layer, the linear layer and the activation layer in sequence to obtain a directional base feature gating unit.

[0059] S205. Establish a graph attention network as an adaptive channel feature fusion network.

[0060] Steps S201-S205 can construct a depression recognition algorithm architecture (DeHTP, Deep Network for House-Tree-Person), so that depression samples can be identified from house-tree-person drawing information.

[0061] The structure and principle of DeHTP are as follows Figure 2 See Figure 2 ,DeHTP consists of three modules: DrawingBasis Encoder, Guided Basis Gated Module and Adaptive Channel Feature Fusion Network.

[0062] Steps S201-S202 are steps for establishing a picture-based feature encoder. In this embodiment, the structure and principle of the established picture-based feature encoder are as follows: Figure 3 See Figure 3 , the painting-based feature encoder is composed of a series of convolutional layers. Except for the first convolutional layer (the first convolutional layer), the remaining convolutional layers form two residual modules and two grouped residual modules. Specifically, the second convolutional layer is a component of the residual module, and the grouped convolutional layer is a component of the grouped residual module.

[0063] Reference Figure 3 In the painting-based feature encoder established by steps S201-S202, the structure of each residual module is the same as the residual block used by the Resnet18 network. The output channels of the two residual modules are 128 and 256 respectively.

[0064] Reference Figure 3 , the grouped residual module is different from the residual module. Specifically, each second convolutional layer in the residual module can be replaced by a grouped convolutional layer, thereby obtaining a grouped residual module.

[0065] Reference Figure 3, the channels in the grouped residual module are divided into k groups, the channels within a group can exchange information between channels, but the channels between groups are not allowed. k is a configurable number.

[0066] Reference Figure 3 The painting-based feature encoder established through steps S201-S202 can extract features from house-tree-person painting information. The output of the painting-based feature encoder is multiple groups of feature maps. In this embodiment, the multiple groups of feature maps output by the painting-based feature encoder are called grouped basis features.

[0067] In this embodiment, by executing steps S201-S202, a classic residual module and an adjusted grouped residual module are used to form a painting-based feature encoder, which can reduce the number of parameters when extracting group-based feature information, save computing resources, and avoid the problem of overfitting.

[0068] Steps S203-S204 are steps for establishing a directional base feature gating unit. In this embodiment, the structure and principle of the established directional base feature gating unit are as follows: Figure 4 shown.

[0069] In order to explore how many sub-rules are needed to judge the information of a house-tree-person painting, we can preset a sufficiently large k value so that Figure 2 The model shown can learn to decide which group rules to use when faced with different paintings during training. Based on the above ideas, steps S203-S204 are designed to establish a directional base feature gating unit.

[0070] Reference Figure 4 , the guided base feature gating unit can finally output a group gate information (Grouped ChannelGate) as a mask to filter out unnecessary sub-rules. Figure 4 The guided base feature gating unit includes a global average pooling layer (Global Average Pooling), a linear layer (Linear, also known as a fully connected layer, etc.) and an activation layer (Sigmoid), etc.

[0071] Step S205 is a step of establishing an adaptive channel feature fusion network. In this embodiment, the structure and principle of the adaptive channel feature fusion network established are as follows: Figure 5 See Figure 5 ,The main structure of the adaptive channel feature fusion network is a graph attention network (Graph Attention Network).

[0072] In this embodiment, when executing step S4, that is, inputting each group of base feature information into the directional base feature gating unit for processing, and obtaining the group gate information and multiple basis vectors output by the directional base feature gating unit, the following steps can be specifically performed:

[0073] S401. Process each group of basic feature information using a global average pooling layer to obtain the basis vectors corresponding to each group of basic feature information;

[0074] S402. Process each basis vector using a linear layer to obtain a query vector corresponding to each basis vector;

[0075] S403. Obtain the key vector corresponding to each basis vector;

[0076] S404. The query vectors are used to form a query matrix, and the key vectors are used to form a key matrix;

[0077] S405. Determine the group attention vector according to the query matrix and the key matrix;

[0078] S406. Processing the group attention vector using an activation layer;

[0079] S407. Process the output result of the activation layer according to a preset threshold to obtain group gate information.

[0080] In step S401, refer to Figure 4 The group basis feature information of size h×w×c output by the painting basis feature encoder can be processed by global average pooling in the global average pooling layer in the guided basis feature gating unit to obtain basis vectors representing each group of basis feature information. In this embodiment, each group of basis feature information is processed to obtain a corresponding basis vector.

[0081] In step S402, the basis vectors may be reshaped so that the basis vectors become a matrix of size k×c / k, and then the matrix is ​​processed using a linear layer in the guided basis feature gating unit to obtain query vectors corresponding to each basis vector.

[0082] In step S403, the key vector corresponding to each basis vector may be obtained.

[0083] In step S404, the query vectors corresponding to all basis vectors can form a query matrix W query , and the key vectors corresponding to all basis vectors can form a key matrix W key .

[0084] In step S405, according to the query matrix W query and bond matrix W key , determine the group attention vector (GroupAttention Vector). Specifically, refer to Figure 4 , we can query the matrix W query and bond matrix W key Perform matrix multiplication to obtain a new matrix, and then take the elements on the diagonal of the new matrix (for example, you can take the elements on the diagonal from the upper left corner to the lower right corner) to obtain a new vector, and use this new vector as the group attention vector.

[0085] The i-th element in the group attention vector is the dot product between the query value (query) and the key value (key) of the i-th basis vector, representing the attention score of this basis vector.

[0086] In step S406, the group attention vector is processed using the activation layer (Sigmoid) in the directional base feature gating unit. Then, step S407 is performed, which may specifically perform the following steps:

[0087] S40701. Traverse each element in the output result of the activation layer;

[0088] S40702. For any traversed element, compare the value of the element with a preset threshold;

[0089] S40703. When the value of the element is greater than a preset threshold, the value of the element is assigned to 1, otherwise, the value of the element is assigned to 0.

[0090] When executing steps S40701-S40703, you can first read the first element in the output result of the activation layer, compare the first element with the preset threshold λ, if the value of the first element is greater than the preset threshold λ, then assign the value of the first element to 1, if the value of the first element is less than or equal to the preset threshold λ, then assign the value of the first element to 0; then read the second element in the output result of the activation layer, compare the second element with the preset threshold λ, if the value of the second element is greater than the preset threshold λ, then assign the value of the second element to 1, if the value of the second element is less than or equal to the preset threshold λ, then assign the value of the second element to 0... Repeat the above operation until all elements in the output result of the activation layer are assigned values.

[0091] By executing steps S40701-S40703, a group gate information (Grouped Channel Gate) can be obtained, which can also be called a group channel gate. In this embodiment, the group gate information can be used as a mask to determine whether the basis vector will be input into the adaptive channel feature fusion network.

[0092] In this embodiment, when executing step S5, that is, inputting the group gate information and each basis vector into the adaptive channel feature fusion network for processing, and obtaining the depression recognition result output by the adaptive channel feature fusion network, the following steps can be specifically performed:

[0093] S501. Determine the adjacency matrix according to the group gate information and each basis vector;

[0094] S502. Use the graph attention network to perform inter-group feature fusion processing on the adjacency matrix to obtain depression recognition results.

[0095] In this embodiment, when executing step S501, that is, determining the adjacency matrix according to the group gate information and each basis vector, the following steps may be specifically performed:

[0096] S50101. Filter each basis vector according to the group gate information to determine the filtered basis vector and the retained basis vector;

[0097] S50102. Connect the retained basis vectors to obtain the adjacency matrix.

[0098] In this embodiment, the basis vectors of each group can be regarded as the node features of the graph, and the connection relationship between each node is determined by the group gate information generated by the previous module (i.e., the directional basis feature gating unit). Therefore, the adjacency matrix required for the graph can be constructed through the group gate information. Specifically, when executing step S50101, the group gate information can be used as a mask, and each basis vector can be used as a graph to perform a mask operation on the group gate information and each basis vector. If the result corresponding to a basis vector after the mask operation is different from this basis vector, then this basis vector is a filtered basis vector; if the result corresponding to a basis vector after the mask operation is the same as this basis vector, then this basis vector is a retained basis vector. The filtered basis vector can be regarded as having no connection relationship with other basis vectors (whether they are retained basis vectors or other filtered basis vectors).

[0099] In step S50102, all the retained basis vectors obtained in step S50101 are connected to obtain an adjacency matrix (Ajacent Matrix).

[0100] In step S502, refer to Figure 5, the adjacency matrix obtained by executing steps S50101-S50102 is input into the graph attention network for processing, and the graph attention network performs inter-group feature fusion on the adjacency matrix, thereby realizing channel information interaction of the channel.

[0101] In step S502, refer to Figure 5 , the output results of the graph attention network are averaged by node features, and input into the linear layer for linearization and other processing, and finally the depression recognition result is output. The depression recognition result indicates whether the painter (the identified person) of the house-tree-person painting information has depression or not.

[0102] In this embodiment, Figure 2 The DeHTP shown can first complete the training process and then be used for depression recognition after training.

[0103] Specifically, during the training process, steps S1-S5 may also be performed, and the house-tree-person drawing information obtained in step S1 may be a hand-drawn drawing of a person to be identified who is known to have depression. In order to train DeHTP, a house-tree-person drawing dataset may be constructed, for example, data may be collected from healthy people and people with confirmed depression; specifically, more than 100 hand-drawn house-tree-person drawing information of confirmed depressed patients and more than 100 hand-drawn house-tree-person drawing information of a healthy control group may be collected, together with the patient's overview of the overall painting and the main elements of the house-tree-person. For example, in one practice, 123 hand-drawn house-tree-person drawing information of confirmed depressed patients and 105 hand-drawn house-tree-person drawing information of a healthy control group may be collected, forming a house-tree-person drawing dataset containing 228 house-tree-person drawing information in total. Each house-tree-person drawing information in the house-tree-person drawing dataset is marked with its corresponding actual depression situation, for example, for the hand-drawn house-tree-person drawing information of confirmed depressed patients, the marked actual depression situation is "with depression"; for example, for the hand-drawn house-tree-person drawing information of a healthy control group, the marked actual depression situation is "without depression".

[0104] During the training process, steps S1-S5 are executed for each house-tree-person painting information, so as to obtain the depression recognition result (actual output result) in step S5, and the actual depression situation corresponding to the house-tree-person painting information is used as the expected output, and the loss function value is calculated according to the actual output result and the expected output, and the network parameters of at least one module in the painting base feature encoder, the directional base feature gating unit and the adaptive channel feature fusion network are adjusted according to the loss function value, so as to realize the training of DeHTP.

[0105] For example, here are the relevant hyperparameters for a training run:

[0106] The number of samples in the healthy control group and the number of samples in the case group for training was 105, and the number of samples in the case group was 123. Data augmentation was performed using horizontal rotation, vertical flipping, and both. The number of groups k used in the group convolution in DeHTP was set to 8, and the threshold used in the guided base feature gating unit was 0.5. The number of group base features of each group output by the painting base feature encoder was 128, and similarly, the length of the node feature vector of the graph constructed later was also 128. The number of layers in the graph attention network was 2, and each layer was dropped with a probability of 0.2 before input. The model performance was evaluated using a 10-fold cross-validation technique. Adam was used as the optimizer with a learning rate of 0.01 and a weight decay of 0.00001. The number of rounds was 400. DeHTP was trained and tested on two NVIDIA Tesla v100s.

[0107] During the depression identification process, steps S1-S5 may be executed. At this time, the house-tree-person drawing information obtained in step S1 may be a hand-drawn drawing of a person to be identified whose identity is unknown. The depression identification result obtained in step S5 indicates a prediction of whether the person to be identified suffers from depression.

[0108] In the process of depression identification, the depression identification results obtained can be used for auxiliary diagnosis of depression and can also be used for purposes unrelated to medical treatment. For example, for a house-tree-person painting information, it can be processed by executing steps S1-S5 to obtain the corresponding depression identification results, and the depression identification results can be used as the emotional color classification information of the house-tree-person painting information, and the house-tree-person painting information can be marked, and the house-tree-person painting information can be used for operations such as sharing. The house-tree-person painting information can be searched through the emotional color classification information, so as to be used for artistic creation under the corresponding emotional color requirements.

[0109] A computer program that executes the depression identification method based on house-tree-person drawing and deep learning in this embodiment can be written and written into a computer device or storage medium. When the computer program is read out and executed, the depression identification method based on house-tree-person drawing and deep learning in this embodiment is executed, thereby achieving the same technical effect as the depression identification method based on house-tree-person drawing and deep learning in the embodiment.

[0110] It should be noted that, unless otherwise specified, when a feature is referred to as being "fixed" or "connected" to another feature, it may be directly fixed or connected to the other feature, or it may be indirectly fixed or connected to the other feature. In addition, the descriptions of up, down, left, right, etc. used in the present disclosure are only relative to the relative positional relationship of the components of the present disclosure in the accompanying drawings. The singular forms of "a", "" and "the" used in the present disclosure are also intended to include the plural forms, unless the context clearly indicates other meanings. In addition, unless otherwise defined, all technical and scientific terms used in this embodiment have the same meaning as those generally understood by those skilled in the art. The terms used in the specification of this embodiment are only for describing specific embodiments and are not intended to limit the present invention. The term "and / or" used in this embodiment includes any combination of one or more related listed items.

[0111] It should be understood that, although the term first, second, third etc. may be adopted to describe various elements in the present disclosure, these elements should not be limited to these terms. These terms are only used to distinguish the same type of elements from each other. For example, without departing from the scope of the present disclosure, the first element may also be referred to as the second element, and similarly, the second element may also be referred to as the first element. The use of any and all examples or exemplary language ("for example", "such as" etc.) provided by the present embodiment is only intended to better illustrate embodiments of the present invention, and unless otherwise required, the scope of the present invention will not be limited.

[0112] It should be appreciated that embodiments of the present invention may be implemented or enforced by computer hardware, a combination of hardware and software, or by computer instructions stored in a non-transitory computer-readable memory. The method may be implemented in a computer program using standard programming techniques - including a non-transitory computer-readable storage medium configured with a computer program, wherein the storage medium so configured causes the computer to operate in a specific and predefined manner - according to the methods and drawings described in the specific embodiments. Each program may be implemented in a high-level procedural or object-oriented programming language to communicate with a computer system. However, if desired, the program may be implemented in assembly or machine language. In any case, the language may be a compiled or interpreted language. In addition, the program may be run on a programmed dedicated integrated circuit for this purpose.

[0113] In addition, the operations of the process described in this embodiment may be performed in any suitable order, unless otherwise indicated in this embodiment or otherwise clearly contradicted by the context. The process described in this embodiment (or variations and / or combinations thereof) may be performed under the control of one or more computer systems configured with executable instructions, and may be implemented as a code (e.g., executable instructions, one or more computer programs, or one or more applications) executed on one or more processors in common, by hardware or a combination thereof. A computer program includes a plurality of instructions that may be executed by one or more processors.

[0114] Further, the method can be implemented in any type of computing platform that is operably connected to a suitable computer, including but not limited to a personal computer, a minicomputer, a mainframe, a workstation, a network or distributed computing environment, a separate or integrated computer platform, or in communication with a charged particle tool or other imaging device, etc. Various aspects of the present invention can be implemented in machine-readable code stored on a non-transitory storage medium or device, whether removable or integrated into a computing platform, such as a hard disk, an optical read and / or write storage medium, a RAM, a ROM, etc., so that it can be read by a programmable computer, and when the storage medium or device is read by the computer, it can be used to configure and operate the computer to perform the process described herein. In addition, the machine-readable code, or part thereof, can be transmitted via a wired or wireless network. When such media includes instructions or programs that implement the above steps in conjunction with a microprocessor or other data processor, the invention of this embodiment includes these and other different types of non-transitory computer-readable storage media. When programmed according to the methods and techniques of the present invention, the present invention also includes the computer itself.

[0115] The computer program can be applied to input data to perform the functions of the present embodiment, thereby converting the input data to generate output data stored in a non-volatile memory. The output information can also be applied to one or more output devices such as a display. In a preferred embodiment of the present invention, the converted data represents a physical and tangible object, including a specific visual depiction of the physical and tangible object produced on the display.

[0116] The above are only preferred embodiments of the present invention. The present invention is not limited to the above embodiments. As long as the technical effects of the present invention are achieved by the same means, any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present invention shall be included in the scope of protection of the present invention. Within the scope of protection of the present invention, its technical solutions and / or implementation methods may have various modifications and changes.

Claims

1. A depression identification method based on house-tree-person drawing and deep learning, characterized in that: The depression identification method based on house-tree-person drawing and deep learning includes: Get house-tree-person drawing information; Obtaining a painting-based feature encoder, a directional-based feature gating unit, and an adaptive channel feature fusion network; Inputting the house-tree-person painting information into the painting-based feature encoder for processing, and obtaining a plurality of group-based feature information output by the painting-based feature encoder; Input each of the group basis feature information into the directional basis feature gating unit for processing, and obtain the group gate information and multiple basis vectors output by the directional basis feature gating unit; wherein each of the basis vectors corresponds to a corresponding one of the group basis feature information; The group gate information and each of the basis vectors are input into the adaptive channel feature fusion network for processing, and a depression recognition result output by the adaptive channel feature fusion network is obtained.

2. The depression identification method based on house-tree-person drawing and deep learning according to claim 1 is characterized in that: The method of obtaining a painting base feature encoder, a directional base feature gating unit and an adaptive channel feature fusion network includes: Establishing a first convolutional layer, at least two residual modules and at least two grouped residual modules; wherein the residual module includes a plurality of second convolutional layers, and the grouped residual module includes a plurality of grouped convolutional layers; The first convolutional layer, each of the residual modules and each of the grouped residual modules are connected in sequence to obtain the painting-based feature encoder.

3. The depression identification method based on house-tree-person drawing and deep learning according to claim 1 is characterized in that: The method of obtaining a painting base feature encoder, a directional base feature gating unit and an adaptive channel feature fusion network includes: Create global average pooling layers, linear layers, and activation layers; The global average pooling layer, the linear layer and the activation layer are connected in sequence to obtain the directional base feature gating unit.

4. The depression identification method based on house-tree-person drawing and deep learning according to claim 1 is characterized in that: The method of obtaining a painting base feature encoder, a directional base feature gating unit and an adaptive channel feature fusion network includes: A graph attention network is established as the adaptive channel feature fusion network.

5. The depression identification method based on house-tree-person drawing and deep learning according to claim 3 is characterized in that: The step of inputting each of the group basis feature information into the directional basis feature gating unit for processing, and obtaining the group gate information and a plurality of basis vectors output by the directional basis feature gating unit, comprises: Using the global average pooling layer to process each of the group basis feature information, to obtain the basis vector corresponding to each of the group basis feature information; Using the linear layer to process each of the basis vectors, to obtain a query vector corresponding to each of the basis vectors; Obtaining the key vectors corresponding to the basis vectors; The query vectors are used to form a query matrix, and the key vectors are used to form a key matrix; Determining a group attention vector based on the query matrix and the key matrix; Processing the group of attention vectors using the activation layer; The output result of the activation layer is processed according to a preset threshold to obtain the group gate information.

6. The depression identification method based on house-tree-person drawing and deep learning according to claim 5 is characterized in that: The step of processing the output result of the activation layer according to a preset threshold to obtain the group gate information includes: Traversing each element in the output result of the activation layer; For any traversed element, compare the value of the element with the preset threshold; When the value of the element is greater than the preset threshold, the value of the element is assigned to 1, otherwise, the value of the element is assigned to 0.

7. The depression identification method based on house-tree-person drawing and deep learning according to claim 4 is characterized in that: The step of inputting the group gate information and each of the basis vectors into the adaptive channel feature fusion network for processing, and obtaining a depression recognition result output by the adaptive channel feature fusion network, comprises: Determine an adjacency matrix according to the group gate information and each of the basis vectors; The graph attention network is used to perform inter-group feature fusion processing on the adjacency matrix to obtain the depression recognition result.

8. The depression identification method based on house-tree-person drawing and deep learning according to claim 7 is characterized in that: The step of determining an adjacency matrix according to the group gate information and each of the basis vectors includes: Filtering each of the basis vectors according to the group gate information to determine the filtered basis vectors and the retained basis vectors; The retained basis vectors are connected to obtain the adjacency matrix.

9. A computer device, characterized in that: It includes a memory and a processor, the memory is used to store at least one program, and the processor is used to load at least one program to execute the depression identification method based on house-tree-person drawing and deep learning as described in any one of claims 1-8.

10. A computer-readable storage medium storing a program executable by a processor, characterized in that: The program executable by the processor is used to execute the depression identification method based on house-tree-person drawing and deep learning as described in any one of claims 1 to 8 when executed by the processor.

Images