Image feature extraction method, device, electronic device, and storage medium

Abstract

The present invention provides an image feature extraction method, device, electronic device, and storage medium, the method comprising: obtaining N feature images of different sizes of a target image, N being an integer greater than 2; using the N images of different sizes as inputs of at least one double-layer bidirectional feature pyramid neural network module; obtaining the output of the at least one double-layer bidirectional feature pyramid neural network module as the image feature of the target image, so as to perform target detection on the target image. The method and device provided by the present invention improve the smoothness of information flow in BiFPN, so as to better integrate features at different levels, improve the expressiveness of features, and thus improve the performance of image processing functions such as target detection and image classification.

Description

Background Art

[0002] Feature pyramids are an essential component in recognition systems for detecting objects at different scales. The inherent multi-scale pyramid hierarchy of deep convolutional networks can be exploited to construct feature pyramids with marginal extra loss. Currently, a top-down architecture with lateral connections is designed to build high-level semantic feature maps at all scales. This architecture, called Feature Pyramid Network (FPN), shows significant improvements over plain feature extractors in several applications.

[0003] There are currently many variants of the feature pyramid network, such as Path Aggregation Network (PANet), simplified PANet, NAS-FPN, BiSFPN, etc. BiFPN is a neural network module in EfficientDet (target detection algorithm series) that fuses features at different levels. BiFPN draws on the structural characteristics of PANet, adds a feature channel from bottom to top, shortens the information transmission process, and thus changes the effect of feature fusion. At the same time, BiFPN uses the stacking of multiple feature pyramids to further improve the effect of feature fusion.

[0004] However, the current planar structure formed by the stacking of BiFPN is from top to bottom, then from bottom to top, and then repeated continuously. Because there is a sequence relationship between from top to bottom and from bottom to top, this planar structure will cause poor information transmission.

[0005] Therefore, how to improve the smooth flow of information in BiFPN to better integrate features at different levels and improve the expressiveness of features, thereby improving target detection performance, is a technical problem that needs to be urgently solved by those skilled in the art. Summary of the invention

[0006] In order to overcome the defects of the above-mentioned prior art, the present invention provides an image feature extraction method, device, electronic device, and storage medium to improve the smooth flow of information in BiFPN, to better integrate features at different levels, to improve the expression ability of features, and thus to improve target detection performance.

[0007] According to one aspect of the present invention, there is provided an image feature extraction method, comprising:

[0008] Obtain N feature images of different sizes of the target image, where N is an integer greater than 2;

[0009] Taking N images of different sizes as input to at least one two-layer bidirectional feature pyramid neural network module;

[0010] The output of the at least one two-layer bidirectional feature pyramid neural network module is obtained as the image feature of the target image to perform target detection on the target image.

[0011] In some embodiments of the present invention, the two-layer bidirectional feature pyramid neural network module includes:

[0012] N input nodes;

[0013] A first intermediate layer, the first intermediate layer includes N-2 groups of nodes, wherein one node in each group of nodes in the first intermediate layer forms a channel from top to bottom, and another node in each group of nodes in the first intermediate layer forms a channel from bottom to top;

[0014] A second intermediate layer, the second intermediate layer includes N groups of nodes, wherein one node in each group of nodes in the second intermediate layer forms a channel from bottom to top, and another node in each group of nodes in the second intermediate layer forms a channel from top to bottom;

[0015] N output nodes, each output node is obtained by fusing a group of nodes of the corresponding second intermediate layer.

[0016] In some embodiments of the present invention, in the first intermediate layer:

[0017] A node X of the i-th group of nodes i1 A node X from the i-1th group of nodes i-11 and the i+1th input node IN i+1 Fusion obtains another node X of the i-th group of nodes i2 From another node X in the i+1th group of nodes i-12 and the i+1th input node IN i+1 The fusion is obtained, wherein i is an integer greater than 1 and less than N-2, and N is an integer greater than 4.

[0018] In some embodiments of the present invention, in the first intermediate layer:

[0019] A node X in the first set of nodes 11 The first input node IN1 and the second input node IN2 are fused to obtain another node X in the first group of nodes. 12 The second input node IN2 and another node X of the second group of nodes 22 Fusion acquisition;

[0020] A node X of the N-2th group of nodes N-21 A node X from the N-3 group of nodes N-31 and the N-1th input node IN N-1 Fusion obtains another node X of the N-2 group of nodesN-22 By the N-1th input node IN N-1 and the Nth input node IN N Obtained by fusion.

[0021] In some embodiments of the present invention, in the second intermediate layer:

[0022] A node Y of the i-th group of nodes i1 By the i-th input node IN i , a node X of the i-1th group of nodes in the first intermediate layer i-11 and a node Y of the i+1th group of nodes i+11 Fusion obtains another node Y of the i-th group of nodes i2 By the i-th input node IN i , another node X of the i-1th group of nodes in the first intermediate layer i-12 and another node Y of the i-1th group of nodes i-12 Obtained by fusion.

[0023] In some embodiments of the present invention, in the second intermediate layer:

[0024] A node Y of the first set of nodes 11 The first input node IN1 and a node Y of the second group of nodes 21 Fusion obtains another node Y of the first group of nodes 12 The first input node IN1 and another node X of the first group of nodes of the first intermediate layer 12 Fusion acquisition;

[0025] A node Y of the Nth group of nodes N1 By the Nth input node IN N and a node X of the N-2th group of nodes in the first intermediate layer N-21 Fusion obtains another node Y of the Nth group of nodes N2 By the Nth input node IN N and another node Y of the N-1th group of nodes N-12 Obtained by fusion.

[0026] In some embodiments of the present invention, when N images of different sizes are used as inputs of multiple double-layer bidirectional feature pyramid neural network modules, the output of the previous double-layer bidirectional feature pyramid neural network module is used as the input of the next double-layer bidirectional feature pyramid neural network module.

[0027] According to another aspect of the present invention, there is also provided an image feature extraction device, comprising:

[0028] A first acquisition module is used to obtain N feature images of different sizes of the target image, where N is an integer greater than 2;

[0029] An input module, used for taking N images of different sizes as inputs of at least one two-layer bidirectional feature pyramid neural network module;

[0030] An output module is used to obtain the output of the at least one two-layer bidirectional feature pyramid neural network module as the image feature of the target image to perform target detection on the target image.

[0031] According to another aspect of the present invention, there is also provided an electronic device, comprising: a processor; and a storage medium storing a computer program, wherein the computer program executes the steps of the image feature extraction method described above when executed by the processor.

[0032] According to another aspect of the present invention, a storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the steps of the image feature extraction method described above are executed.

[0033] Compared with the prior art, the advantages of the present invention are:

[0034] By expanding the planar feature pyramid in BiFPN into a three-dimensional feature fusion method, the expansion method is to transform the planar feature pyramid network into two planes, one plane is first from top to bottom and then from bottom to top, and the other plane is first from bottom to top and then from top to bottom, and then fuse the outputs of the two planes, and finally stack these two planes to form a new pyramid network structure, which solves the problem of the planar structure caused by the top-to-bottom and bottom-to-top channel sequence and stacking of the feature pyramid in BiFPN, and prevents the efficiency of information flow or fusion. Therefore, the image feature extraction method and device can be applied to image processing functions such as target detection and image classification. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] The above and other features and advantages of the present invention will become more apparent by describing in detail exemplary embodiments thereof with reference to the attached drawings.

[0036] Figure 1 A flow chart of an image feature extraction method according to an embodiment of the present invention is shown.

[0037] Figure 2 Schematic diagram of a BiFPN module showing five input nodes.

[0038] Figure 3 A schematic diagram of a double-layer bidirectional feature pyramid neural network module according to an embodiment of the present invention is shown.

[0039] Figure 4 A module diagram of an image feature extraction device according to an embodiment of the present invention is shown.

[0040] Figure 5 A schematic diagram of a computer-readable storage medium in an exemplary embodiment of the present disclosure is schematically shown.

[0041] Figure 6 A schematic diagram of an electronic device in an exemplary embodiment of the present disclosure is schematically shown. DETAILED DESCRIPTION

[0042] Example embodiments will now be described more fully with reference to the accompanying drawings. However, example embodiments can be implemented in a variety of forms and should not be construed as limited to the examples set forth herein; rather, these embodiments are provided so that the disclosure will be more comprehensive and complete and to fully convey the concepts of the example embodiments to those skilled in the art. The described features, structures, or characteristics may be combined in any suitable manner in one or more embodiments.

[0043] In addition, the accompanying drawings are only schematic illustrations of the present disclosure and are not necessarily drawn to scale. The same reference numerals in the figures represent the same or similar parts, and their repeated description will be omitted. Some of the block diagrams shown in the accompanying drawings are functional entities and do not necessarily correspond to physically or logically independent entities. These functional entities can be implemented in software form, or implemented in one or more hardware modules or integrated circuits, or implemented in different networks and / or processor devices and / or microcontroller devices.

[0044] In order to solve the defects of the prior art, the present invention provides an image feature extraction method, device, electronic device, and storage medium by simplifying the production of moving pictures to achieve batch generation while avoiding text errors caused by manual operation.

[0045] See first Figure 1 , Figure 1 A schematic diagram of an image feature extraction method according to an embodiment of the present invention is shown. The image feature extraction method comprises the following steps:

[0046] Step S110: Obtain N feature images of different sizes of the target image, where N is an integer greater than 2.

[0047] Specifically, the target image can be sequentially passed through the convolutional neural network to obtain feature images of different sizes. Further, the number of feature images obtained sequentially can be greater than or equal to N, so that the last N feature images obtained can be selected in step S110. Among the N-radius images, the later the feature images are obtained, the smaller the size is.

[0048] Step S120: using N images of different sizes as inputs to at least one two-layer bidirectional feature pyramid neural network module.

[0049] Specifically, the two-layer bidirectional feature pyramid neural network module includes: N input nodes, a first intermediate layer, a second intermediate layer and N output nodes. The first intermediate layer includes N-2 groups of nodes, wherein one node in each group of nodes in the first intermediate layer forms a channel from top to bottom, and another node in each group of nodes in the first intermediate layer forms a channel from bottom to top. The second intermediate layer includes N groups of nodes, wherein one node in each group of nodes in the second intermediate layer forms a channel from bottom to top, and another node in each group of nodes in the second intermediate layer forms a channel from top to bottom. Each output node is obtained by fusion of a corresponding group of nodes in the second intermediate layer. In the following, Figure 3 The specific structure of the double-layer bidirectional feature pyramid neural network module provided by the present invention is described and will not be elaborated here.

[0050] Specifically, when N images of different sizes are used as inputs of multiple double-layer bidirectional feature pyramid neural network modules, the output of the previous double-layer bidirectional feature pyramid neural network module is used as the input of the next double-layer bidirectional feature pyramid neural network module. In this way, multiple double-layer bidirectional feature pyramid neural network modules can be cascaded horizontally.

[0051] Step S130: Obtain the output of the at least one two-layer bidirectional feature pyramid neural network module as the image feature of the target image to perform target detection on the target image.

[0052] In the image feature extraction method provided by the present invention, the planar feature pyramid in BiFPN is expanded into a three-dimensional feature fusion method. The expansion method is to convert the planar feature pyramid network into two planes, one plane is first from top to bottom and then from bottom to top, and the other plane is first from bottom to top and then from top to bottom. The outputs of the two planes are then fused, and finally the two planes are stacked to form a new pyramid network structure, which solves the problem of the planar structure caused by the sequence of channels from top to bottom and from bottom to top in the feature pyramid in BiFPN and the stacking, and prevents the efficiency of information flow or fusion.

[0053] See below Figure 2 , Figure 2 Schematic diagram of a BiFPN module showing five input nodes. Figure 2Five input nodes, three first intermediate nodes, five second intermediate nodes and five output nodes are shown. Among them, the first intermediate node X1 is obtained by merging the input node In1 and the input node In2. The first intermediate node X2 is obtained by merging the input node In3 and the first intermediate node X1. The first intermediate node X3 is obtained by merging the input node In4 and the first intermediate node X2. Thus, a channel from top to bottom is formed. The second intermediate node Y5 is obtained by merging the input node In5 and the first intermediate node X3. The second intermediate node Y4 is obtained by merging the input node In4, the first intermediate node X3 and the second intermediate node Y5. The second intermediate node Y3 is obtained by merging the input node In3, the first intermediate node X2 and the second intermediate node Y4. The second intermediate node Y2 is obtained by merging the input node In2, the first intermediate node X1 and the second intermediate node Y3. The second intermediate node Y1 is obtained by merging the input node In1 and the second intermediate node Y2. Thus, a channel from bottom to top is formed. Finally, five output nodes are output through five second intermediate nodes. Therefore, the planar structure formed by the stacking of BiFPN is from top to bottom, then from bottom to top, and then repeated continuously, because there is a sequence relationship between from top to bottom and from bottom to top. Therefore, this planar structure will cause poor information transmission, thereby affecting the performance of target detection.

[0054] See below Figure 3 , Figure 3 A schematic diagram of a double-layer bidirectional feature pyramid neural network module according to an embodiment of the present invention is shown.

[0055] In the present invention, in the first intermediate layer, a node X of the i-th group of nodes i1 A node X from the i-1th group of nodes i-11 and the i+1th input node IN i+1 Fusion obtains another node X of the i-th group of nodes i2 From another node X in the i+1th group of nodes i-12 and the i+1th input node IN i+1 Fusion is obtained, wherein i is an integer greater than 1 and less than N-2, and N is an integer greater than 4. At the same time, in the first intermediate layer: a node X of the first group of nodes 11 The first input node IN1 and the second input node IN2 are fused to obtain another node X in the first group of nodes. 12 The second input node IN2 and another node X of the second group of nodes 22 Fusion obtains: a node X of the N-2 group of nodes N-21 A node X from the N-3 group of nodes N-31 and the N-1th input node IN N-1 Fusion obtains another node X of the N-2 group of nodes N-22By the N-1th input node IN N-1 and the Nth input node IN N In the present invention, in the second intermediate layer, a node Y of the i-th group of nodes i1 By the i-th input node IN i , a node X of the i-1th group of nodes in the first intermediate layer i-11 and a node Y of the i+1th group of nodes i+11 Fusion obtains another node Y of the i-th group of nodes i2 By the i-th input node IN i , another node X of the i-1th group of nodes in the first intermediate layer i-12 and another node Y of the i-1th group of nodes i-12 At the same time, in the second intermediate layer, a node Y of the first group of nodes 11 The first input node IN1 and a node Y of the second group of nodes 21 Fusion obtains another node Y of the first group of nodes 12 The first input node IN1 and another node X of the first group of nodes of the first intermediate layer 12 Fusion obtains; a node Y of the Nth group of nodes N1 By the Nth input node IN N and a node X of the N-2th group of nodes in the first intermediate layer N-21 Fusion obtains another node Y of the Nth group of nodes N2 By the Nth input node IN N and another node Y of the N-1th group of nodes N-12 The structure of the above two-layer bidirectional feature pyramid neural network module is as follows: Figure 3 As shown, no further description is given here.

[0056] In various embodiments of the present invention, the two-layer bidirectional feature pyramid neural network module can have 3 inputs, 4 inputs, or 5 inputs, and the present invention is not limited thereto. The downward channel will be upsampled, the upward channel will be downsampled, and the parallel input may not have a sampling operation. The node input will be added first, and then the information can be fused through an activated convolution layer that does not change the height, width, and depth of the feature map.

[0057] The above is merely an illustrative description of multiple implementations of the present invention, and the present invention is not limited thereto.

[0058] The present invention also provides an image feature extraction device. Figure 4The schematic diagram of the image feature extraction device according to the embodiment of the present invention is shown. The image feature extraction device 200 comprises a first acquisition module 210 , an input module 220 , and an output module 230 .

[0059] The first acquisition module 210 is used to obtain N feature images of different sizes of the target image, where N is an integer greater than 2.

[0060] The input module 220 is used to use N images of different sizes as inputs of at least one two-layer bidirectional feature pyramid neural network module.

[0061] The output module 230 is used to obtain the output of the at least one two-layer bidirectional feature pyramid neural network module as the image feature of the target image to perform target detection on the target image.

[0062] In the image feature extraction device provided by the present invention, the planar feature pyramid in BiFPN is expanded into a three-dimensional feature fusion method. The expansion method is to convert the planar feature pyramid network into two planes, one plane is first from top to bottom and then from bottom to top, and the other plane is first from bottom to top and then from top to bottom. The outputs of the two planes are then fused, and finally the two planes are stacked to form a new pyramid network structure, which solves the problem of the planar structure caused by the sequence of channels from top to bottom and from bottom to top in the feature pyramid in BiFPN and the stacking, and prevents the efficiency of information flow or fusion.

[0063] Figure 4 The image feature extraction device provided by the present invention is only schematically shown. Without violating the concept of the present invention, the splitting, merging and adding of modules are all within the protection scope of the present invention. The image feature extraction device provided by the present invention can be implemented by software, hardware, firmware, plug-ins and any combination thereof, and the present invention is not limited thereto.

[0064] In an exemplary embodiment of the present disclosure, a computer-readable storage medium is also provided, on which a computer program is stored, and when the program is executed by, for example, a processor, the steps of the image feature extraction method described in any of the above embodiments can be implemented. In some possible implementations, various aspects of 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 steps of various exemplary embodiments of the present invention described in the above image feature extraction method section of this specification.

[0065] refer to Figure 5As shown, a program product 400 for implementing the above method according to an embodiment of the present invention is described, which can adopt a portable compact disk read-only memory (CD-ROM) and include program code, and can be run on a terminal device, such as a personal computer. However, the program product of the present invention is not limited thereto. In this document, a readable storage medium can be any tangible medium containing or storing a program, which can be used by or in combination with an instruction execution system, an apparatus or a device.

[0066] The program product may use any combination of one or more readable media. The readable medium may be a readable signal medium or a readable storage medium. The readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or device, or any combination of the above. More specific examples (non-exhaustive list) of readable storage media include: an electrical connection with one or more wires, a portable disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above.

[0067] The computer readable storage medium may include a data signal propagated in a baseband or as part of a carrier wave, wherein a readable program code is carried. This propagated data signal may take a variety of forms, including but not limited to an electromagnetic signal, an optical signal, or any suitable combination of the above. The readable storage medium may also be any readable medium other than a readable storage medium, which may send, propagate, or transmit a program for use by an instruction execution system, an apparatus, or a device or used in combination with it. The program code contained on the readable storage medium may be transmitted with any appropriate medium, including but not limited to wireless, wired, optical cable, RF, etc., or any suitable combination of the above.

[0068] Program code for performing the operations of the present invention may be written in any combination of one or more programming languages, including object-oriented programming languages ​​such as Java, C++, etc., and conventional procedural programming languages ​​such as "C" or similar programming languages. The program code may be executed entirely on the tenant computing device, partially on the tenant device, as a separate software package, partially on the tenant computing device and partially on a remote computing device, or entirely on a remote computing device or server. In cases involving remote computing devices, the remote computing device may be connected to the tenant computing device via any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computing device (e.g., via the Internet using an Internet service provider).

[0069] In an exemplary embodiment of the present disclosure, an electronic device is also provided, which may include a processor and a memory for storing executable instructions of the processor, wherein the processor is configured to execute the steps of the image feature extraction method in any of the above embodiments by executing the executable instructions.

[0070] It will be appreciated by those skilled in the art that various aspects of the present invention may be implemented as a system, method or program product. Therefore, various aspects of the present invention may be specifically implemented in the following forms, namely: a complete hardware implementation, a complete software implementation (including firmware, microcode, etc.), or a combination of hardware and software, which may be collectively referred to herein as a "circuit", "module" or "system".

[0071] Refer to the following Figure 6 The electronic device 600 according to this embodiment of the present invention is described. Figure 6 The electronic device 600 shown is only an example and should not bring any limitation to the functions and scope of use of the embodiments of the present invention.

[0072] like Figure 6 As shown, the electronic device 600 is in the form of a general computing device. The components of the electronic device 600 may include, but are not limited to: at least one processing unit 610, at least one storage unit 620, a bus 630 connecting different system components (including the storage unit 620 and the processing unit 610), a display unit 640, etc.

[0073] The storage unit stores a program code, which can be executed by the processing unit 610, so that the processing unit 610 performs the steps according to various exemplary embodiments of the present invention described in the above image feature extraction method section of this specification. For example, the processing unit 610 can perform the following steps: Figure 1 to Figure 2 Follow the steps shown in .

[0074] The storage unit 620 may include a readable medium in the form of a volatile storage unit, such as a random access memory unit (RAM) 6201 and / or a cache memory unit 6202 , and may further include a read-only memory unit (ROM) 6203 .

[0075] The storage unit 620 may also include a program / utility 6204 having a set (at least one) of program modules 6205, such program modules 6205 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 the implementation of a network environment.

[0076] Bus 630 may represent one or more of several types of bus structures, including a memory unit bus or memory unit controller, a peripheral bus, an accelerated graphics port, a processing unit, or a local bus using any of a variety of bus architectures.

[0077] The electronic device 600 may also communicate with one or more external devices 700 (e.g., keyboards, pointing devices, Bluetooth devices, etc.), may communicate with one or more devices that enable tenants to interact with the electronic device 600, and / or may communicate with any device that enables the electronic device 600 to communicate with one or more other computing devices (e.g., routers, modems, etc.). Such communication may be performed through an input / output (I / O) interface 650. In addition, the electronic device 600 may 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) through a network adapter 660. The network adapter 660 may communicate with other modules of the electronic device 600 through a bus 630. It should be understood that, although not shown in the figure, other hardware and / or software modules may be used in conjunction with the electronic device 600, including but not limited to: microcode, device drivers, redundant processing units, external disk drive arrays, RAID systems, tape drives, and data backup storage systems.

[0078] Through the description of the above implementation, it is easy for those skilled in the art to understand that the example implementation described here can be implemented by software, or by software combined with necessary hardware. Therefore, the technical solution according to the implementation of the present disclosure can be embodied in the form of a software product, which can be stored in a non-volatile storage medium (which can be a CD-ROM, a USB flash drive, a mobile hard disk, etc.) or on a network, including a number of instructions to enable a computing device (which can be a personal computer, a server, or a network device, etc.) to execute the above-mentioned image feature extraction method according to the implementation of the present disclosure.

[0079] Compared with the prior art, the advantages of the present invention are:

[0080] By expanding the planar feature pyramid in BiFPN into a three-dimensional feature fusion method, the expansion method is to transform the planar feature pyramid network into two planes, one plane is first from top to bottom and then from bottom to top, and the other plane is first from bottom to top and then from top to bottom, and then the outputs of the two planes are fused, and finally these two planes are stacked to form a new pyramid network structure, which solves the problem of the planar structure caused by the top-to-bottom and bottom-to-top channel sequence and stacking of the feature pyramid in BiFPN, which prevents the efficiency of information flow or fusion.

[0081] Those skilled in the art will readily appreciate other embodiments of the present disclosure after considering the specification and practicing the invention disclosed herein. This application is intended to cover any modification, use or adaptation of the present disclosure, which follows the general principles of the present disclosure and includes common knowledge or customary techniques in the art that are not disclosed in the present disclosure. The specification and examples are intended to be exemplary only, and the true scope and spirit of the present disclosure are indicated by the appended claims.

Claims

1. A method for extracting image features, characterized in that: include: Obtain N feature images of different sizes of the target image, where N is an integer greater than 2; Taking N images of different sizes as inputs of at least one double-layer bidirectional feature pyramid neural network module, the double-layer bidirectional feature pyramid neural network module comprises: N input nodes; A first intermediate layer, the first intermediate layer includes N-2 groups of nodes, wherein one node in each group of nodes in the first intermediate layer forms a channel from top to bottom, and another node in each group of nodes in the first intermediate layer forms a channel from bottom to top; A second intermediate layer, the second intermediate layer includes N groups of nodes, wherein one node in each group of nodes in the second intermediate layer forms a channel from bottom to top, and another node in each group of nodes in the second intermediate layer forms a channel from top to bottom; N output nodes, each of which is obtained by fusing a group of nodes in the corresponding second intermediate layer; The output of the at least one two-layer bidirectional feature pyramid neural network module is obtained as the image feature of the target image to perform target detection on the target image.

2. The image feature extraction method according to claim 1, characterized in that: In the first intermediate layer: A node X of the i-th group of nodes i1 A node X from the i-1th group of nodes (i-1)1 and the i+1th input node IN i+1 Fusion obtains another node X of the i-th group of nodes i2 From another node X in the i+1th group of nodes (i-1)2 and the i+1th input node IN i+1 The fusion is obtained, wherein i is an integer greater than 1 and less than N-2, and N is an integer greater than 4.

3. The image feature extraction method according to claim 2, characterized in that: In the first intermediate layer: A node X in the first set of nodes 11 The first input node IN1 and the second input node IN2 are fused to obtain another node X in the first group of nodes. 12 The second input node IN2 and another node X of the second group of nodes 22 Fusion acquisition; A node X of the N-2 group of nodes (N-2)1 A node X from the N-3 group of nodes (N-3)1 and the N-1th input node IN N-1 Fusion obtains another node X of the N-2 group of nodes (N-2)2 By the N-1th input node IN N-1 and the Nth input node IN N Obtained by fusion.

4. The image feature extraction method according to claim 3, characterized in that: In the second intermediate layer: A node Y of the i-th group of nodes i1 By the i-th input node IN i , a node X of the i-1th group of nodes in the first intermediate layer (i-1)1 and a node Y of the i+1th group of nodes (i+1)1 Fusion obtains another node Y of the i-th group of nodes i2 By the i-th input node IN i , another node X of the i-1th group of nodes in the first intermediate layer (i-1)2 and another node Y of the i-1th group of nodes (i-1)2 Obtained by fusion.

5. The image feature extraction method according to claim 4, characterized in that: In the second intermediate layer: A node Y of the first set of nodes 11 The first input node IN1 and a node Y of the second group of nodes 21 Fusion obtains another node Y of the first group of nodes 12 The first input node IN1 and another node X of the first group of nodes of the first intermediate layer 12 Fusion acquisition; A node Y of the Nth group of nodes N1 By the Nth input node IN N and a node X of the N-2th group of nodes in the first intermediate layer (N-2)1 Fusion obtains another node Y of the Nth group of nodes N2 By the Nth input node IN N and another node Y of the N-1th group of nodes (N-1)2 Obtained by fusion.

6. The image feature extraction method according to any one of claims 1 to 5, characterized in that: When N images of different sizes are used as inputs of multiple double-layer bidirectional feature pyramid neural network modules, the output of the previous double-layer bidirectional feature pyramid neural network module is used as the input of the next double-layer bidirectional feature pyramid neural network module.

7. An image feature extraction device, characterized in that: include: A first acquisition module is used to obtain N feature images of different sizes of the target image, where N is an integer greater than 2; An input module is used to use N images of different sizes as inputs of at least one double-layer bidirectional feature pyramid neural network module, wherein the double-layer bidirectional feature pyramid neural network module includes: N input nodes; A first intermediate layer, the first intermediate layer includes N-2 groups of nodes, wherein one node in each group of nodes in the first intermediate layer forms a channel from top to bottom, and another node in each group of nodes in the first intermediate layer forms a channel from bottom to top; A second intermediate layer, the second intermediate layer includes N groups of nodes, wherein one node in each group of nodes in the second intermediate layer forms a channel from bottom to top, and another node in each group of nodes in the second intermediate layer forms a channel from top to bottom; N output nodes, each of which is obtained by fusing a group of nodes in the corresponding second intermediate layer; An output module is used to obtain the output of the at least one two-layer bidirectional feature pyramid neural network module as the image feature of the target image to perform target detection on the target image.

8. An electronic device, characterized in that: The electronic device comprises: processor; A storage medium having a computer program stored thereon, wherein the computer program, when executed by the processor, executes the image feature extraction method according to any one of claims 1 to 6.

9. A storage medium, characterized in that: The storage medium stores a computer program, and when the computer program is executed by a processor, the image feature extraction method according to any one of claims 1 to 6 is executed.

Images