CN117392409A - Work clothes identification method, device, equipment and computer readable storage medium - Google Patents

Work clothes identification method, device, equipment and computer readable storage medium Download PDF

Info

Publication number
CN117392409A
CN117392409A CN202210775655.0A CN202210775655A CN117392409A CN 117392409 A CN117392409 A CN 117392409A CN 202210775655 A CN202210775655 A CN 202210775655A CN 117392409 A CN117392409 A CN 117392409A
Authority
CN
China
Prior art keywords
image
work clothes
target work
target
saliency
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Pending
Application number
CN202210775655.0A
Other languages
Chinese (zh)
Inventor
马佳炯
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
SF Technology Co Ltd
Original Assignee
SF Technology Co Ltd
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by SF Technology Co Ltd filed Critical SF Technology Co Ltd
Priority to CN202210775655.0A priority Critical patent/CN117392409A/en
Publication of CN117392409A publication Critical patent/CN117392409A/en
Pending legal-status Critical Current

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/40—Extraction of image or video features
    • G06V10/46—Descriptors for shape, contour or point-related descriptors, e.g. scale invariant feature transform [SIFT] or bags of words [BoW]; Salient regional features
    • G06V10/462—Salient features, e.g. scale invariant feature transforms [SIFT]
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • G06N3/084—Backpropagation, e.g. using gradient descent
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
    • G06V10/74—Image or video pattern matching; Proximity measures in feature spaces
    • G06V10/761—Proximity, similarity or dissimilarity measures
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
    • G06V10/77—Processing image or video features in feature spaces; using data integration or data reduction, e.g. principal component analysis [PCA] or independent component analysis [ICA] or self-organising maps [SOM]; Blind source separation
    • G06V10/80—Fusion, i.e. combining data from various sources at the sensor level, preprocessing level, feature extraction level or classification level
    • G06V10/806—Fusion, i.e. combining data from various sources at the sensor level, preprocessing level, feature extraction level or classification level of extracted features
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
    • G06V10/82—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using neural networks

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Evolutionary Computation (AREA)
  • General Physics & Mathematics (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Health & Medical Sciences (AREA)
  • Artificial Intelligence (AREA)
  • Computing Systems (AREA)
  • Software Systems (AREA)
  • General Health & Medical Sciences (AREA)
  • Multimedia (AREA)
  • Medical Informatics (AREA)
  • Databases & Information Systems (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Biomedical Technology (AREA)
  • Biophysics (AREA)
  • Computational Linguistics (AREA)
  • Data Mining & Analysis (AREA)
  • Molecular Biology (AREA)
  • General Engineering & Computer Science (AREA)
  • Mathematical Physics (AREA)
  • Image Analysis (AREA)

Abstract

The embodiment of the application provides a method, a device, equipment and a computer readable storage medium for identifying work clothes, wherein the method comprises the following steps: acquiring a target work clothes image to be identified; performing significance detection on the target work clothes image to obtain a significance area image in the target work clothes image; and determining a work clothes recognition result of the target work clothes image according to the first image feature of the target work clothes image and the second image feature of the saliency area image. According to the work clothes identification method, the saliency detection is carried out on the target work clothes image, the saliency area image in the target work clothes image is extracted, the image characteristics of the target work clothes image and the saliency area image are comprehensively utilized, the target work clothes image can be analyzed on different scales such as the whole scale and the local scale, and therefore the accuracy of work clothes identification results of the work clothes image is effectively improved, and the work clothes misjudgment rate is reduced.

Description

Work clothes identification method, device, equipment and computer readable storage medium
Technical Field
The embodiment of the application relates to the technical field of image recognition, in particular to a method, a device, equipment and a computer readable storage medium for identifying work clothes.
Background
In some financial scenes, when some businesses are transacted, in order to ensure safety, information of clients often needs to be checked. In order to reduce the workload of bank staff, an online auditing function is realized at present, namely, a client can automatically shoot and upload according to specified requirements through specific mobile phone software, and a background server automatically judges whether the uploaded pictures meet corresponding business requirements.
For some businesses that require specific enterprise personnel to transact, the backend server often makes decisions by identifying the work clothes on the client. However, since the work clothes of the large enterprises are often solid-colored, the work clothes are similar on the whole, so that the accuracy of identifying the work clothes at the background service end is not high at present, and a high misjudgment rate exists.
Disclosure of Invention
The embodiment of the application provides a method, a device, equipment and a computer readable storage medium for identifying work clothes, which aim to solve the technical problems that the accuracy rate of identifying the work clothes is not high and a certain misjudgment rate exists in the prior art.
In one aspect, an embodiment of the present application provides a method for identifying a work service, including:
acquiring a target work clothes image to be identified;
Performing significance detection on the target work clothes image to obtain a significance area image in the target work clothes image;
and determining a work clothes recognition result of the target work clothes image according to the first image feature of the target work clothes image and the second image feature of the saliency area image.
As a possible embodiment of the present application, the performing saliency detection on the target work clothes image to obtain a saliency area image in the target work clothes image includes:
inputting the target work clothes image into a trained work clothes identification model for processing, and outputting a center coordinate and a length and a width;
and extracting a salient region image in the target work clothes image according to the center coordinates and the length and width.
As another possible embodiment of the present application, the processing the inputting the target work clothes image into the trained work clothes identification model, outputting the center coordinate and the length and width includes:
inputting the target work clothes image into an image block coding layer in a trained work clothes identification model for processing, and outputting an image coding vector of the target work clothes image;
inputting a preset initial region detection vector and the image coding vector into a self-attention coding layer in the work clothes recognition model for processing, and outputting a salient region detection vector;
And inputting the salient region detection vector into an attention module in the work clothes identification model, and outputting a center coordinate and a length and a width.
As another possible embodiment of the present application, the performing saliency detection on the target work clothes image to obtain a saliency area image in the target work clothes image includes:
performing significance detection on the target work clothes image to obtain a first area image in the target work clothes image;
if the resolution of the target work clothes image is higher than a preset resolution threshold, performing significance detection on the first area image to obtain a second area image in the first area image;
and setting the first area image and the second area image as saliency area images in the target work clothes image.
As another possible embodiment of the present application, the determining, according to the first image feature of the target work clothes image and the second image feature of the salient region image, the work clothes identification result of the target work clothes image includes:
fusing the first image features of the target work clothes image and the second image features of the saliency area image to obtain fused image features;
Calculating the similarity between the fusion image features and preset candidate work service features;
determining target work clothes features from the candidate work clothes features according to the magnitude relation of the similarity;
and setting the work clothes type corresponding to the target work clothes characteristic as a work clothes identification result of the target work clothes image.
As another possible embodiment of the present application, before determining the work clothes recognition result of the target work clothes image according to the first image feature of the target work clothes image and the second image feature of the saliency area image, the method includes:
inputting the preset initial image characteristics and the image coding vectors of the target work clothes image into a self-attention coding layer in a trained work clothes identification model for processing, and outputting the image characteristic vectors;
inputting preset initial image features and image coding vectors of the salient region images into a self-attention coding layer in the work clothes identification model for processing, and outputting region feature vectors;
and respectively inputting the image feature vector and the region feature vector into a convolution layer in the work clothes recognition model for processing, and outputting a first image feature and a second image feature.
As another possible embodiment of the present application, before the performing the saliency detection on the target working clothes image to obtain a saliency area image in the target working clothes image, the method further includes:
performing edge detection on the target work clothes image to obtain key points of the target work clothes image;
determining the proportion of a first area and a second area in the target work service image according to the position information of the key points;
and if the proportion is smaller than a preset proportion threshold value, executing the step of performing saliency detection on the target work clothes image to obtain a saliency area image in the target work clothes image.
On the other hand, the embodiment of the application also provides a work clothes identification device, which comprises:
the acquisition module is used for acquiring a target work clothes image to be identified;
the detection module is used for carrying out saliency detection on the target work clothes image to obtain a saliency area image in the target work clothes image;
and the identification module is used for determining the work clothes identification result of the target work clothes image according to the first image characteristic of the target work clothes image and the second image characteristic of the saliency area image.
On the other hand, the embodiment of the application also provides a work clothes identification device, which comprises a processor, a memory and a work clothes identification program stored in the memory and capable of running on the processor, wherein the processor executes the work clothes identification program to realize the steps in the work clothes identification method.
In another aspect, an embodiment of the present application further provides a computer readable storage medium, where a work clothes identification program is stored, where the work clothes identification program is executed by a processor to implement the steps in the work clothes identification method described above.
According to the work clothes identification method, the saliency detection is carried out on the target work clothes image, the saliency area image in the target work clothes image is extracted, the image characteristics of the target work clothes image and the saliency area image are comprehensively utilized, the target work clothes image can be analyzed on different scales such as the whole scale and the local scale, and therefore the accuracy of work clothes identification results of the work clothes image is effectively improved, and the work clothes misjudgment rate is reduced.
Drawings
In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings that are needed in the description of the embodiments will be briefly described below, it being obvious that the drawings in the following description are only some embodiments of the present invention, and that other drawings may be obtained according to these drawings without inventive effort for a person skilled in the art.
Fig. 1 is a schematic diagram of an implementation scenario of a method for identifying a work service according to an embodiment of the present application;
fig. 2 is a schematic step flow diagram of a method for identifying work clothes according to an embodiment of the present application;
FIG. 3 is a schematic flowchart of a step of significance detection according to an embodiment of the present disclosure;
fig. 4 is a schematic flowchart of a step of implementing saliency detection based on a network model according to an embodiment of the present application;
FIG. 5 is a flowchart illustrating another exemplary significance detection procedure according to an embodiment of the present disclosure;
fig. 6 is a flowchart illustrating steps for determining a work service recognition result based on image features according to an embodiment of the present application;
fig. 7 is a flowchart illustrating a step of extracting image features according to an embodiment of the present application;
FIG. 8 is a flowchart illustrating steps of processing an image by a work clothes recognition model according to an embodiment of the present application;
fig. 9 is a flowchart of a step of performing work clothes identification based on an edge detection result of a target work clothes image according to an embodiment of the present application;
fig. 10 is a schematic structural diagram of a work clothes identification device according to an embodiment of the present application;
fig. 11 is a schematic structural diagram of a work clothes identification device according to an embodiment of the present application.
Detailed Description
The following description of the technical solutions in the embodiments of the present application will be made clearly and completely with reference to the drawings in the embodiments of the present application, and it is apparent that the described embodiments are only some embodiments of the present invention, not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be encompassed by the present invention.
In the embodiments of the present application, the term "exemplary" is used to mean "serving as an example, instance, or illustration. Any embodiment described as "exemplary" in this application is not necessarily to be construed as preferred or advantageous over other embodiments. The following description is presented to enable any person skilled in the art to make and use the invention. In the following description, details are set forth for purposes of explanation. It will be apparent to one of ordinary skill in the art that the present invention may be practiced without these specific details. In other instances, well-known structures and processes have not been described in detail so as not to obscure the description of the invention with unnecessary detail. Thus, the present invention is not intended to be limited to the embodiments shown, but is to be accorded the widest scope consistent with the principles and features disclosed in the embodiments of the present application.
The embodiment of the application provides a method, a device, equipment and a computer readable storage medium for identifying work clothes, and the detailed description is given below.
The work clothes identification method in the embodiment of the application is that the work clothes identification device is deployed on the work clothes identification device in a program mode, the work clothes identification device is installed in the work clothes identification equipment in a processor mode, and after the work clothes identification device in the work clothes identification equipment acquires a target work clothes image to be identified, the following steps are executed by running the program corresponding to the work clothes identification method: performing saliency detection on the target work clothes image to obtain a saliency area image in the target work clothes image, determining a work clothes identification result of the target work clothes image according to the first image characteristic of the target work clothes image and the second image characteristic of the saliency area image, and outputting a final work clothes identification result.
As shown in fig. 1, fig. 1 is a schematic view of an implementation scenario of a method for identifying a work piece according to an embodiment of the present application, where the implementation scenario provided in the embodiment of the present application includes an image acquisition module 100 and a work piece identification device 200 communicatively connected to the image acquisition module 100. The image acquisition module 100 is mainly used for acquiring an image of a target work piece to be identified and sending the image to the work piece identification device 200. Specifically, the image acquisition module may be a user terminal, such as a mobile phone, or a software program deployed on the user terminal, such as corresponding APP software. The work clothes identification device 200 is generally integrally disposed in a server, and the server may be, for example, a cloud server or a server existing in other forms, which is not described herein.
It should be noted that, the schematic implementation scenario of the work clothes identification method shown in fig. 1 is only an example, and the implementation scenario of the work clothes identification method described in the embodiment of the present application is for more clearly describing the technical solution of the embodiment of the present application, and does not constitute a limitation to the technical solution provided in the embodiment of the present application.
Based on the implementation scene diagram of the work clothes identification method, a specific embodiment of the work clothes identification method is provided.
As shown in fig. 2, fig. 2 is a schematic step flow diagram of a method for identifying a work service according to an embodiment of the present application, where the method for identifying a work service in the embodiment of the present application includes steps 201 to 203:
and 201, acquiring a target work clothes image to be identified.
Considering that the embodiment of the present application is mainly used for implementing the identification of the work clothes, the target work clothes image to be identified generally refers to an image containing the self-dressing, i.e. the work clothes information, which is shot by a handheld mobile terminal, such as a mobile phone, when a customer handles a financial service that needs a specific enterprise employee to handle. After the corresponding image is shot, the mobile terminal transmits the image to a rear end server of the software program, namely the work clothes identification device through the software program deployed on the mobile terminal, so that the work clothes identification device takes the image as a target work clothes image to be identified. Of course, it should be noted that the target work image may also include other information required for handling the financial service, such as face information, hand-held certificate information, etc. The embodiments of the present application are not described herein.
And 202, performing significance detection on the target work clothes image to obtain a significance area image in the target work clothes image.
In this embodiment of the present application, the saliency area image refers to an area image with saliency, that is, an area image of interest of a user, where in general, the saliency area image may include a mark identifier of saliency, for example, a relatively common enterprise logo, an employee identity identifier, and so on.
In the embodiment of the application, the saliency area image in the target work clothes image can be obtained by performing saliency detection on the target work clothes image. There are many algorithms for performing saliency detection specifically, for example, saliency detection is relatively commonly implemented based on the contrast of pixels in an image, that is, the distance between pixel values of each pixel point, and saliency detection can also be implemented based on frequency information of the image. However, considering that the effect of the significance detection is not ideal enough, as an alternative embodiment of the present application, an implementation scheme for implementing the significance detection based on a neural network model of deep learning is proposed, and particularly, reference may be made to fig. 3 and the explanation thereof.
As another optional embodiment of the present application, to further improve the subsequent recognition effect, if the resolution of the collected target work clothes image is relatively high, the significance detection may be performed on the target work clothes image under multiple scales to obtain a significant area image under multiple different scales, and at this time, a specific implementation scheme may refer to the subsequent fig. 5 and the explanation content thereof.
Considering that in the process of identifying the target work clothes image, in order to avoid that the user in the target work clothes image is not just opposite to shooting and the significant area of the work clothes is blocked, deformed and the like, so as to influence the subsequent work clothes identification effect, as a further alternative embodiment of the application, before the significance detection is performed on the target work clothes image, edge detection is performed on the target work clothes image, so as to determine whether the key point in the target work clothes image is just opposite to shooting, and a specific implementation scheme can refer to the following fig. 9 and the content of explanation thereof.
203, determining a work clothes recognition result of the target work clothes image according to the first image feature of the target work clothes image and the second image feature of the saliency area image.
In the embodiment of the application, after the saliency area image in the target work clothes image is acquired, the target work clothes image can be analyzed from different scales such as the whole scale and the local scale based on the image characteristics of the target work clothes image on the whole and the local, so that the work clothes identification result of the target work clothes image is obtained.
Specifically, there are various implementations of determining the work clothes recognition result of the target work clothes image according to the image characteristics. For example, as a possible embodiment of the present application, the classification model obtained through deep learning thought training may be used to process the image features of the target work clothes image, so as to output the probability that the target work clothes image belongs to the preset multiple work clothes types, and then the work clothes type with the highest probability is used as the work clothes recognition result of the target work clothes image.
However, the work clothes recognition result of the target work clothes image through the classification model often needs to define a plurality of possible work clothes types in advance, namely when the target work clothes image containing the new work clothes type needs to be recognized, a new classification model needs to be trained again, that is, the applicability of the work clothes recognition result of the target work clothes image through the classification model is poor. Based on this, another possible embodiment of the present application is provided, by calculating the similarity between the image feature of the target work clothes image and the image feature of the preset candidate work clothes image, and setting the work clothes type corresponding to the candidate work clothes image with the highest similarity as the work clothes recognition result of the target work clothes image, further, in order to ensure the work clothes recognition effect of the implementation scheme, in the process of extracting the image feature of the target work clothes image, it is required to ensure that the similarity between the image features extracted from the work clothes images belonging to the same work clothes is higher, and the similarity between the image features extracted from the work clothes images not belonging to the same work clothes is lower, so that the neural network model with the effect can be obtained based on the idea training of deep learning, and the specific implementation scheme can refer to the following fig. 6 and the explanation content thereof.
According to the work clothes identification method, the saliency detection is carried out on the target work clothes image, the saliency area image in the target work clothes image is extracted, the image characteristics of the target work clothes image and the saliency area image are comprehensively utilized, the target work clothes image can be analyzed on different scales such as the whole scale and the local scale, and therefore the accuracy of work clothes identification results of the work clothes image is effectively improved, and the work clothes misjudgment rate is reduced.
As shown in fig. 3, fig. 3 is a schematic flowchart of a step of significance detection provided in an embodiment of the present application, which is described in detail below.
In the embodiment of the application, an implementation scheme for realizing significance detection through the deep learning idea is provided, specifically, the implementation scheme comprises steps 301-302:
and 301, inputting the target work clothes image into a trained work clothes identification model for processing, and outputting a center coordinate and a length and a width.
In this embodiment of the present application, the trained work clothes recognition model refers to a model that is obtained by training based on the idea of deep learning by using a large number of sample images, and the specific training process is not described herein in detail. The specific structure of the work clothes recognition model may be constructed based on the processing procedure of the model on the image, specifically, as an alternative embodiment of the application, the model may firstly encode the image, and then determine the position information of the salient region from the encoded image based on the attention mechanism, where the position information of the salient region may be composed of a central coordinate and a range, that is, length and width information, and at this time, the specific procedure of the model on the image processing may refer to the content of fig. 4 and the explanation thereof.
And 302, extracting a salient region image in the target work clothes image according to the center coordinates and the length and width.
In the embodiment of the application, after the position information of the salient region in the target work clothes image is determined, the salient region image in the target work clothes image can be extracted. Specifically, taking the center coordinate and the length and width as the position information as an example, if the lower left corner of the target work clothes image is taken as the origin to establish a coordinate system, at this time, the value range of the abscissa of the saliency area image in the target work clothes image is half of the length of the abscissa in the center coordinate, and the value range of the ordinate is half of the length of the ordinate in the center coordinate.
As shown in fig. 4, fig. 4 is a schematic flowchart of steps for implementing saliency detection based on a network model according to an embodiment of the present application, and specifically includes steps 401 to 403:
and 401, inputting the target work clothes image into an image block coding layer in the trained work clothes identification model for processing, and outputting an image coding vector of the target work clothes image.
In this embodiment of the present application, before inputting the target work clothes image into the image block coding layer in the trained work clothes recognition model for processing, the target work clothes image is cut into a plurality of blocks, for example, the target work clothes image may be cut into 4×4, 8×8 or 16×16 blocks, where the blocks may or may not overlap. Further, each block is input to the image block encoding layer for processing, and is mapped into a high-dimensional feature vector, for example 768-dimensional or 1024-dimensional. Specifically, taking the example of cutting the target work service picture into 16 blocks of 4×4, at this time, the 16 image blocks are subjected to coding processing to obtain feature vectors of 16×768 dimensions, and the feature vectors are the image coding vectors of the target work service image.
And 402, inputting a preset initial region detection vector and the image coding vector into a self-attention coding layer in the work clothes recognition model for processing, and outputting a salient region detection vector.
In this embodiment, after obtaining the image coding vector of the target working clothes image, a preset initial region detection vector is additionally added on the basis of the image coding vector to form a 17 x 768-dimensional vector, and then the image coding vector obtained by fusion is input into a self-attention coding layer in the working clothes recognition model, and at this time, the self-attention coding layer fully considers the association relationship between the input different feature vectors, namely, the association relationship between the initial region detection vector and the feature vectors of each image block. Considering that the output vector of the self-attention encoding layer is the same as the input vector in dimension, that is, a 17×768-dimensional output vector is finally output, the output vector includes a 768-dimensional saliency region detection vector and a 16×768-dimensional processed image encoding vector, wherein the output vector can be used for representing the position information of the saliency region in the target working image.
403, inputting the saliency area detection vector into an attention module in the work clothes identification model, and outputting a center coordinate and a length and a width.
In this embodiment of the present application, after obtaining the saliency area detection vector for representing the position information of the saliency area in the target work clothes image, the saliency area detection vector is input to the attention module in the work clothes recognition model again to perform analysis processing, so that a final four-dimensional vector can be output, and the value of each dimension in the four-dimensional vector represents the abscissa and the ordinate of the center coordinate, and the length and the width of the saliency area.
Further, as an optional embodiment of the present application, the salient region detection vector and the processed image coding vector are input to the attention module in the work clothes recognition model, where the salient region detection vector and the information contained in the image coding vector can be fused, so as to obtain more accurate information
As shown in fig. 5, fig. 5 is a schematic flowchart of another step of significance detection provided in the embodiment of the present application, which is described in detail below.
In this embodiment of the present application, saliency detection is performed on a target work clothes image under a plurality of different scales based on the resolution of the target work clothes image, and specifically includes steps 501 to 504:
and 501, performing significance detection on the target work clothes image to obtain a first area image in the target work clothes image.
In the embodiment of the application, the first area image in the target work clothes image can be obtained by performing significance detection on the target work clothes image. Specific implementation may refer to the foregoing step 202 and the implementation of fig. 3, which are not described herein,
502, judging whether the resolution of the target work clothes image is higher than a preset resolution threshold. If yes, go to step 503; if not, executing other steps.
In the embodiment of the application, the resolution of the target work clothes image and the preset resolution threshold are compared, so that the size relation between the resolution of the target work clothes image and the preset resolution threshold can be obtained. Specifically, if the resolution of the target work clothes image is higher, that is, higher than a preset resolution threshold, multiple saliency detection can be performed on the target work clothes image, so that as many image areas under different scales as possible are acquired, and a more accurate recognition result is obtained, otherwise, if the resolution of the target work clothes image is lower, that is, lower than the preset resolution threshold, at the moment, less information is contained in the image area under the low scale, and better image features cannot be extracted, so that multiple saliency detection is not needed, and a first area image in the target work clothes image is obtained through extraction, wherein the first area image is set as the saliency area image in the target work clothes image.
And 503, performing significance detection on the first region image to obtain a second region image in the first region image.
In this embodiment of the present application, it is known from the foregoing description that if the resolution of the target work clothes image is higher, that is, higher than the preset resolution threshold, multiple saliency detections may be performed on the target work clothes image, so as to obtain as many image areas under different scales as possible. Therefore, the region image obtained by the first saliency detection can be subjected to the second saliency detection, and the saliency region in the first region can be extracted from the first region image again to obtain a second region image.
As another optional embodiment of the present application, the saliency detection may also be performed on the second area image, so as to obtain a third area image in the second area image, where the third area image may also be regarded as a saliency area image in the target work clothes image. The number of times of performing the saliency detection may be related to the resolution of the target work clothes image, and the higher the resolution of the target work clothes image, the more the number of times of performing the saliency detection may be.
And 504, setting the first area image and the second area image as salient area images in the target work clothes image.
In this embodiment of the present application, the first area image may be understood as a salient area image of the target work clothes image at the first scale, and the second area image may be understood as a salient area image of the target work clothes image at the smaller scale, and thus, the first area image and the second area image may be regarded as salient area images in the target work clothes image.
According to the implementation scheme, whether the target work clothes image is subjected to multiple saliency detection or not is judged according to the size relation between the resolution of the target work clothes image and the preset resolution threshold, when the original resolution is higher, the richer image features are obtained through multiple saliency detection, so that the accuracy of work clothes identification is improved, when the original resolution is lower, multiple saliency detection is avoided, the salient region image containing less information is extracted, and the efficiency of work clothes identification is improved.
Fig. 6 is a schematic flowchart of steps for determining a work clothes recognition result based on image features according to an embodiment of the present application, which is described in detail below.
In this embodiment of the present application, specifically, the method determines the recognized result of the work clothes through the similarity between the image features, and specifically includes steps 601 to 604:
And 601, fusing the first image feature of the target work clothes image and the second image feature of the saliency area image to obtain a fused image feature.
In this embodiment of the present application, the first image feature and the second image feature are obtained through the same model processing, so that dimensions of the first image feature and the second image feature are the same. Further, in order to facilitate the subsequent calculation of the similarity, the dimension of the obtained fused image feature should be always the same as the dimension of the preset candidate work service feature, so here, fusing the first image feature and the second image feature generally means that the average value of the first image feature and the second image feature is obtained, and the fused image feature is obtained.
Of course, if the salient region of the target work clothes image under more scales is extracted, more image features can be obtained through the same model processing, and the dimensions of the image features are the same, so that the image features can be fused in an average value mode to obtain a fused image feature with consistent output dimensions all the time.
And 602, calculating the similarity between the fused image features and the preset candidate work service features.
In this embodiment of the present application, the preset candidate work clothes features are image features obtained by performing the same steps as those described above, that is, saliency detection and feature fusion on some given images including the specified work clothes types, where the candidate work clothes features are usually completed in an offline state and stored in a database in advance, and when the target work clothes image needs to be identified, the work clothes identification device reads all the candidate work clothes features from the database and calculates the similarity with the fused image features one by one.
In this embodiment of the present application, the similarity between the fused image feature and the candidate work service feature may be calculated by adopting a cosine similarity calculation manner, and of course, other manners may also be adopted, for example, the similarity between the image features may be determined according to the euclidean distance between the image features, and the rule for specifically calculating the similarity is not limited in this embodiment of the present application.
Furthermore, it should be noted that, in conjunction with the foregoing description, the image features are obtained by processing the image by using a model obtained through specific training, and the similarity between the image features can be used to describe the possibility that the image contains the same type of workpiece, that is, the similarity between the image processed by using the model obtained through specific training and the feature processed by the image containing the same type of workpiece is higher, and the similarity between the image processed by using the image containing different types of workpiece is higher. Specifically, in view of the fact that in the embodiment of the present application, the saliency detection of the image is also implemented based on the neural network model trained by using the idea of deep learning, so, to simplify the flow, the neural network model for extracting the image features may be fused with the model for implementing the saliency detection, that is, the saliency detection of the image and the feature extraction of the image are implemented by using the same model, and at this time, a specific implementation scheme may refer to fig. 7 and the content of explanation thereof.
603, determining a target work clothes feature from the candidate work clothes features according to the magnitude relation of the similarity.
As can be seen in conjunction with the foregoing description, the similarity between image features may be used to describe the likelihood that the same type of work item is contained in an image. Specifically, if the similarity between the fused image feature and a certain candidate work clothes feature is higher, the image corresponding to the target work clothes image and the candidate work clothes feature is more likely to contain the same work clothes type, so that the candidate work clothes feature with the highest similarity can be used as the target work clothes feature.
And 604, setting the work clothes type corresponding to the target work clothes characteristic as a work clothes identification result of the target work clothes image.
In this embodiment, as can be seen from the foregoing description, for each candidate work clothes feature, the candidate work clothes feature is obtained by processing a predetermined image including a specified work clothes type, and the specified work clothes type is generally used as a corresponding work clothes type of the candidate work clothes feature, for example, a work clothes of a certain department of a certain company. Therefore, after the target work feature is determined from the candidate work features, the work type corresponding to the target work feature is set as the work recognition result of the target work image.
Fig. 7 is a schematic flowchart of a step of extracting image features according to an embodiment of the present application, which is described in detail below.
In the embodiment of the present application, in order to simplify the flow, the extraction of the image features is synchronously implemented by using a model for implementing the saliency detection, and specifically includes steps 701 to 703:
701, inputting the preset initial image characteristics and the image coding vectors of the target work clothes image into a self-attention coding layer in a trained work clothes identification model for processing, and outputting the image characteristic vectors.
In this embodiment of the present application, the image coding vector of the target work clothes image is also obtained by inputting the target work clothes image segmented into a plurality of blocks into the image block coding layer in the work clothes identification model for processing, and the specific implementation process is not described herein. Reference is made in particular to step 401 described above and to what has been explained.
Different from the implementation flow for performing saliency detection on an image, in the embodiment of the present application, task identifiers for implementing image feature extraction, that is, preset initial image features and image coding vectors of a target work clothes image, are input to a self-attention coding layer in a trained work clothes recognition model to be processed, so that the self-attention coding layer fully considers the association relationship between the input initial image features and the image coding vectors of the target work clothes image, and thus, image feature vectors related to the target work clothes image are output.
Specifically, taking the foregoing embodiment as an example, if the image coding vector of the target working image is represented by a feature vector of 16×768 dimensions, the preset initial image feature is also a feature vector of 768 dimensions, the preset initial image feature and the image coding vector of the target working image are input into the self-attention coding layer in the trained working recognition model for processing, and since the output vector of the self-attention coding layer is the same as the input vector in dimensions, a 17×768-dimension output vector is finally output, and the output vector includes an 768-dimension image feature vector and a 16×768-dimension processed image coding vector.
And 702, inputting the preset initial image characteristics and the image coding vectors of the salient region images into a self-attention coding layer in the work clothes identification model for processing, and outputting region characteristic vectors.
In this embodiment of the present application, the specific implementation process obtained by inputting the significant area image segmented into a plurality of blocks into the image block coding layer in the work clothes identification model for processing the significant area image coding vector is not described herein. Reference is made in particular to step 401 described above and to what has been explained.
After obtaining the image coding vector of the salient region image, it can be known by combining the related description of the foregoing step 701 that the task identifier for implementing image feature extraction, that is, the preset initial image feature and the image coding vector of the salient region image, are input to the self-attention coding layer in the trained work clothes recognition model for processing, so that the self-attention coding layer fully considers the association relationship between the input initial image feature and the image coding vector of the salient region image, thereby outputting the region feature vector related to the salient region image. Specifically, the regional feature vector is also a 768-dimensional vector.
And 703, respectively inputting the image feature vector and the region feature vector into a convolution layer in the work clothes recognition model for processing, and outputting a first image feature and a second image feature.
In the embodiment of the application, after the image coding vector of the target work clothes image and the image coding vector of the salient region image are processed respectively based on the self-attention mechanism, namely by utilizing the self-attention coding layer, the corresponding feature vector is obtained, the image feature vector and the region feature vector are further input into a convolution layer in the work clothes recognition model to be processed, and the 768-dimensional image feature vector and the region feature vector are expanded into image features with higher dimensionality so as to enrich the information contained in the image features. Specifically, the image feature vector is input into the convolution layer in the work clothes identification model for processing, so that the first image feature of the target work clothes image can be obtained, and the region feature vector is input into the convolution layer in the work clothes identification model for processing, so that the second image feature of the saliency region image can be obtained.
In order to facilitate understanding of the structure of the work clothes recognition model and the complete implementation flow of processing the target work clothes image by the work clothes recognition model provided by the embodiment of the application, as shown in fig. 8, a schematic diagram of the step flow of processing the image by the work clothes recognition model is provided by the embodiment of the application.
In this embodiment of the present application, the work clothes recognition model mainly includes an image block coding layer, a self-attention coding layer, a convolution layer, and an attention module. Specifically, in the training process, parameters in the image block coding layer, the self-attention coding layer, the convolution layer and the attention module need to be trained through a large number of marked sample images. The specific training process is as follows:
1) Randomly initializing parameters of each network layer in the engineering clothes identification model;
2) Inputting the sample image into a random initialization industry service identification model, outputting a first image feature and a second image feature, and calculating a loss value of the industry service identification model according to a preset loss function by combining the preset belonging industry service category of the sample image;
3) Judging whether the loss value is smaller than a preset loss threshold value, if not, updating parameters in the initialized work clothes identification model based on a back propagation algorithm to obtain an updated work clothes identification model, inputting a sample image into the updated work clothes identification model to obtain the loss value of the work clothes identification model, and then executing the step 3) again until the loss values of a plurality of continuous times are smaller than the preset loss threshold value, and setting the currently obtained work clothes identification model as a trained work clothes identification model.
Specifically, during the training process, the loss function is mainly composed of a triplet loss function and a discriminant feature loss. The triplet loss function is used to measure whether the inter-class gap is large enough and the intra-class gap is small enough, and the purpose of the triplet loss is to reduce the gap between sample features in the same class as much as possible and to enlarge the gap between sample features in different classes as much as possible. The specific calculation formula is as follows:
L Tri ＝max(0,dis(a,p)-dis(a,n)+margin 1 )
wherein a represents the characteristic representation of the current image output by the current network, dis (.) is a distance measurement function, cosine distance is used in the application, p is a sample which is similar to the characteristic a but has the greatest distance in the samples of the current batch during training, n is a sample which is different from the characteristic a but has the closest distance in the samples of the current batch during training, and margin1 is an over-parameter, and the value range is between 0 and 1.
And discriminative feature loss in order to encourage different scale outputs to focus on different discriminative regions. I.e. to make the difference between the first image feature and the second image feature as large as possible. Specifically, the calculation formula is:
L R ＝max(0,dis(layer 1 ,layer 2 )+margin 2 )
wherein, margin2 is the super parameter of settlement, and the value range is between 0 ~ 1, and layer1 and layer2 represent the characteristic representation of two different scale outputs respectively, namely first image feature and second image feature in this application. Of course, if the resolution of the image is high, resulting in the use of more scale saliency areas in the actual application scenario, the loss function only calculates the discriminant feature loss of the neighboring scales, and adds the discriminant feature loss of each neighboring scale.
As shown in fig. 9, fig. 9 is a schematic flowchart of steps for performing work clothes identification based on an edge detection result of a target work clothes image according to an embodiment of the present application, and specifically includes steps 901 to 904:
and 901, performing edge detection on the target work clothes image to obtain key points of the target work clothes image.
In the embodiment of the application, the edge detection can extract some edge pixel points in the target work clothes image, so that the outline information of some main bodies in the target work clothes image is obtained. Specifically, key points of the target work image are commonly referred to as sleeves, hems, center seams, buttons, zippers, and the like.
As another alternative embodiment of the present application, considering that in the present application, the target work image generally refers to an image that includes a self-wearing garment captured by a customer when the customer handles a financial service that needs a specific enterprise employee to handle, that is, the target work image generally includes face information in addition to work information, and based on the face information, the target work image may also be captured in a state of being right, so that key points of the target work image may also include some key points of a face in the image, such as a left eye corner, a right eye corner, a left mouth corner, and so on.
And 902, determining the proportion of the first area and the second area in the target work clothes image according to the position information of the key points.
In this embodiment of the present application, according to the aforementioned positional information of the key point, mainly the abscissa of the key point in the image, the ratio of the left half area to the right half area in the target work clothes image may be calculated. For example, taking the case that the key points include left and right eyes, left and right mouth, and nose, the first region may be understood as a left face region size, specifically, a difference between an average abscissa of the left eyes and the left mouth and an abscissa of the nose may be calculated, and the second region may be understood as a right face region size, specifically, a difference between an average abscissa of the right eyes and the right mouth and an abscissa of the nose may be calculated. Of course, other calculation methods for determining the ratio of the first area to the second area in the target work service image according to the position information of the key point are also feasible, and the embodiments of the present application are not described herein again.
903, determining that the ratio is less than a preset ratio threshold. If yes, go to step 904; if not, executing other steps.
In this embodiment of the present application, as can be seen from the foregoing description, when the target working image is acquired in the facing state, the left and right regions should be symmetrical, that is, the difference between the sizes of the first region and the second region is relatively small, so that it can be determined whether the target working image is acquired in the facing state through the size relationship between the ratio of the first region and the second region and the preset ratio threshold. Specifically, if the ratio is smaller than the preset ratio threshold, it indicates that the left and right regions in the target work clothes image have high symmetry, and at this time, the target work clothes image is effectively collected, and subsequent steps can be executed. Otherwise, if the ratio is smaller than the preset ratio threshold, the fact that the left area and the right area in the target work clothes image are not symmetrical is indicated, and the remarkable area of the work clothes in the target work clothes image is possibly shielded, deformed and other abnormal conditions can occur, so that the follow-up work clothes identification effect is affected. Therefore, the target work clothes image acquisition is invalid, and a voice prompt can be output to remind a user to acquire again.
And 904, performing significance detection on the target work clothes image to obtain a significance area image in the target work clothes image.
In this embodiment of the present application, a specific implementation step of performing saliency detection on the target work clothes image may refer to the explanation of step 202, which is not described herein.
In order to better implement the work clothes identification method in the embodiment of the application, the embodiment of the application also provides a work clothes identification device based on the work clothes identification method. As shown in fig. 10, fig. 10 is a schematic structural diagram of a work clothes identification device according to an embodiment of the present application. Specifically, the work clothes recognition device includes:
an acquiring module 1001, configured to acquire a target work clothes image to be identified;
the detection module 1002 is configured to perform saliency detection on the target work clothes image, so as to obtain a saliency area image in the target work clothes image;
and the identification module 1003 is configured to determine a work clothes identification result of the target work clothes image according to the first image feature of the target work clothes image and the second image feature of the saliency area image.
In some embodiments of the present application, the detection module includes:
The model processing submodule is used for inputting the target work clothes image into a trained work clothes identification model for processing and outputting a center coordinate and a length and width;
and the region extraction sub-module is used for extracting the salient region image in the target work clothes image according to the center coordinates and the length and width.
In some embodiments of the present application, the model processing submodule includes:
the coding unit is used for inputting the target work clothes image into an image block coding layer in the trained work clothes identification model for processing and outputting an image coding vector of the target work clothes image;
the self-attention processing unit is used for inputting a preset initial region detection vector and the image coding vector into a self-attention coding layer in the work clothes identification model for processing, and outputting a salient region detection vector;
and the attention processing unit is used for inputting the significance region detection vector into an attention module in the work clothes identification model and outputting a center coordinate and a length and a width.
In some embodiments of the present application, the detection module further includes:
the first detection sub-module is used for performing significance detection on the target work clothes image to obtain a first area image in the target work clothes image;
The second detection sub-module is used for performing significance detection on the first area image if the resolution of the target work service image is higher than a preset resolution threshold value, so as to obtain a second area image in the first area image;
and the setting sub-module is used for setting the first area image and the second area image as salient area images in the target work clothes image.
In some embodiments of the present application, the identification module includes:
a fusion sub-module for fusing the first image feature of the target work service image and the second image feature of the saliency area image to obtain a fused image feature
The similarity calculation sub-module is used for calculating the similarity between the fusion image characteristics and the preset candidate work service characteristics;
the screening sub-module is used for determining target work service characteristics from the candidate work service characteristics according to the magnitude relation of the similarity;
and the work clothes type setting sub-module is used for setting the work clothes type corresponding to the target work clothes characteristic as the work clothes identification result of the target work clothes image.
In some embodiments of the present application, the work clothes identification device further includes a feature extraction module, where the feature extraction module includes:
The coding sub-module is used for inputting the preset initial image characteristics and the image coding vectors of the target work clothes image into a self-attention coding layer in the trained work clothes identification model for processing, and outputting the image characteristic vectors;
the self-attention processing sub-module is used for inputting the preset initial image characteristics and the image coding vectors of the salient region images into the self-attention coding layer in the work clothes identification model for processing, and outputting region characteristic vectors;
and the convolution sub-module is used for respectively inputting the image feature vector and the region feature vector into a convolution layer in the work clothes recognition model for processing and outputting a first image feature and a second image feature.
In some embodiments of the present application, the work clothes identification device further includes a judging module, where the judging module includes:
the edge detection sub-module is used for carrying out edge detection on the target work clothes image to obtain key points of the target work clothes image;
the proportion calculation sub-module is used for determining the proportion of the first area and the second area in the target work clothes image according to the position information of the key points;
and the detection module is used for carrying out significance detection on the target work clothes image if the proportion is smaller than a preset proportion threshold value to obtain a significance area image in the target work clothes image.
The embodiment of the application also provides a work clothes identification device, as shown in fig. 11, and fig. 11 is a schematic structural diagram of the work clothes identification device.
The work clothes identification device comprises a memory, a processor and a work clothes identification program which is stored in the memory and can run on the processor, wherein the processor realizes the steps in the work clothes identification method provided by any embodiment of the application when executing the work clothes identification program.
Specifically, the present invention relates to a method for manufacturing a semiconductor device. The work clothes identification device may include one or more processors 1101 of a processing core, one or more memories 1102 of a storage medium, a power supply 1103, and an input unit 1104, among other components. It will be appreciated by those skilled in the art that the work clothes identification device structure shown in fig. 11 is not limiting of the work clothes identification device and may include more or fewer components than shown, or may combine certain components, or may be arranged in different components. Wherein:
the processor 1101 is a control center of the work clothes identification device, connects various parts of the whole work clothes identification device by various interfaces and lines, and performs various functions and processes of the work clothes identification device by running or executing software programs and/or modules stored in the memory 1102 and calling data stored in the memory 1102, thereby performing overall monitoring of the work clothes identification device. Optionally, the processor 1101 may include one or more processing cores; preferably, the processor 1101 may integrate an application processor that primarily handles operating systems, user interfaces, applications, etc., and a modem processor that primarily handles wireless communications. It will be appreciated that the modem processor described above may not be integrated into the processor 1101.
The memory 1102 may be used to store software programs and modules, and the processor 1101 executes various functional applications and data processing by executing the software programs and modules stored in the memory 1102. The memory 1102 may mainly include a storage program area and a storage data area, wherein the storage program area may store an operating system, application programs required for at least one function (such as a sound playing function, an image playing function, etc.), and the like; the storage data area may store data created according to the use of the work clothes identification device, etc. In addition, memory 1102 may include high-speed random access memory, and may also include non-volatile memory, such as at least one magnetic disk storage device, flash memory device, or other volatile solid-state storage device. Accordingly, the memory 1102 may also include a memory controller to provide the processor 1101 with access to the memory 1102.
The work clothes identification device further includes a power supply 1103 for powering the various components, preferably the power supply 1103 can be logically connected to the processor 1101 by a power management system, such that functions of managing charging, discharging, and power consumption are performed by the power management system. The power supply 1103 may also include one or more of any of a direct current or alternating current power supply, recharging system, power failure detection circuit, power converter or inverter, power status indicator, etc.
The work clothes recognition device may further include an input unit 1104, the input unit 1104 being operable to receive input numerical or character information and to generate keyboard, mouse, joystick, optical or trackball signal inputs related to user settings and function control.
Although not shown, the work clothes recognition apparatus may further include a display unit or the like, which is not described herein. In particular, in this embodiment, the processor 1101 in the work clothes identification device loads executable files corresponding to the processes of one or more application programs into the memory 1102 according to the following instructions, and the processor 1101 runs the application programs stored in the memory 1102, so as to implement the steps in the work clothes identification method provided in any embodiment of the present application.
To this end, embodiments of the present application provide a computer-readable storage medium, which may include: read Only Memory (ROM), random access Memory (RAM, random Access Memory), magnetic or optical disk, and the like. The computer readable storage medium stores a work clothes identification program which when executed by a processor implements the steps in the work clothes identification method provided in any embodiment of the present application.
In the foregoing embodiments, the descriptions of the embodiments are focused on, and the portions of one embodiment that are not described in detail in the foregoing embodiments may be referred to in the foregoing detailed description of other embodiments, which are not described herein again.
In the implementation, each unit or structure may be implemented as an independent entity, or may be implemented as the same entity or several entities in any combination, and the implementation of each unit or structure may be referred to the foregoing method embodiments and will not be repeated herein.
The specific implementation of each operation above may be referred to the previous embodiments, and will not be described herein.
The foregoing has described in detail a method for identifying a work service provided by the embodiments of the present application, and specific examples have been applied herein to illustrate the principles and embodiments of the present invention, and the above description of the embodiments is only for aiding in understanding the method and core idea of the present invention; meanwhile, as those skilled in the art will have variations in the specific embodiments and application scope in light of the ideas of the present invention, the present description should not be construed as limiting the present invention.

Claims (10)

1. A method of identifying a work garment, comprising:
Acquiring a target work clothes image to be identified;
performing significance detection on the target work clothes image to obtain a significance area image in the target work clothes image;
and determining a work clothes recognition result of the target work clothes image according to the first image feature of the target work clothes image and the second image feature of the saliency area image.
2. The method for identifying a work clothes according to claim 1, wherein the performing the saliency detection on the target work clothes image to obtain a saliency area image in the target work clothes image includes:
inputting the target work clothes image into a trained work clothes identification model for processing, and outputting a center coordinate and a length and a width;
and extracting a salient region image in the target work clothes image according to the center coordinates and the length and width.
3. The method for identifying the work clothes according to claim 2, wherein inputting the target work clothes image into a trained work clothes identification model for processing, outputting a center coordinate and a length and width, comprises:
inputting the target work clothes image into an image block coding layer in a trained work clothes identification model for processing, and outputting an image coding vector of the target work clothes image;
Inputting a preset initial region detection vector and the image coding vector into a self-attention coding layer in the work clothes recognition model for processing, and outputting a salient region detection vector;
and inputting the salient region detection vector into an attention module in the work clothes identification model, and outputting a center coordinate and a length and a width.
4. The method for identifying a work clothes according to claim 1, wherein the performing the saliency detection on the target work clothes image to obtain a saliency area image in the target work clothes image includes:
performing significance detection on the target work clothes image to obtain a first area image in the target work clothes image;
if the resolution of the target work clothes image is higher than a preset resolution threshold, performing significance detection on the first area image to obtain a second area image in the first area image;
and setting the first area image and the second area image as saliency area images in the target work clothes image.
5. The method of claim 1, wherein determining the work clothes recognition result of the target work clothes image according to the first image feature of the target work clothes image and the second image feature of the saliency area image comprises:
Fusing the first image features of the target work clothes image and the second image features of the saliency area image to obtain fused image features;
calculating the similarity between the fusion image features and preset candidate work service features;
determining target work clothes features from the candidate work clothes features according to the magnitude relation of the similarity;
and setting the work clothes type corresponding to the target work clothes characteristic as a work clothes identification result of the target work clothes image.
6. The method of claim 1, wherein prior to determining the work item identification result of the target work item image based on the first image feature of the target work item image and the second image feature of the saliency area image, the method comprises:
inputting the preset initial image characteristics and the image coding vectors of the target work clothes image into a self-attention coding layer in a trained work clothes identification model for processing, and outputting the image characteristic vectors;
inputting preset initial image features and image coding vectors of the salient region images into a self-attention coding layer in the work clothes identification model for processing, and outputting region feature vectors;
And respectively inputting the image feature vector and the region feature vector into a convolution layer in the work clothes recognition model for processing, and outputting a first image feature and a second image feature.
7. The method for identifying a work piece according to any one of claims 1 to 6, wherein before the performing the saliency detection on the target work piece image to obtain a saliency area image in the target work piece image, the method further comprises:
performing edge detection on the target work clothes image to obtain key points of the target work clothes image;
determining the proportion of a first area and a second area in the target work service image according to the position information of the key points;
and if the proportion is smaller than a preset proportion threshold value, executing the step of performing saliency detection on the target work clothes image to obtain a saliency area image in the target work clothes image.
8. A work clothes identification device, comprising:
the acquisition module is used for acquiring a target work clothes image to be identified;
the detection module is used for carrying out saliency detection on the target work clothes image to obtain a saliency area image in the target work clothes image;
And the identification module is used for determining the work clothes identification result of the target work clothes image according to the first image characteristic of the target work clothes image and the second image characteristic of the saliency area image.
9. A work clothes identification device, characterized in that it comprises a processor, a memory and a work clothes identification program stored in the memory and executable on the processor, the processor executing the work clothes identification program to implement the steps in the work clothes identification method of any one of claims 1 to 7.
10. A computer-readable storage medium, characterized in that the computer-readable storage medium has stored thereon a work clothes identification program that is executed by a processor to implement the steps in the work clothes identification method of any one of claims 1 to 7.
CN202210775655.0A 2022-07-01 2022-07-01 Work clothes identification method, device, equipment and computer readable storage medium Pending CN117392409A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202210775655.0A CN117392409A (en) 2022-07-01 2022-07-01 Work clothes identification method, device, equipment and computer readable storage medium

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202210775655.0A CN117392409A (en) 2022-07-01 2022-07-01 Work clothes identification method, device, equipment and computer readable storage medium

Publications (1)

Publication Number Publication Date
CN117392409A true CN117392409A (en) 2024-01-12

Family

ID=89461856

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202210775655.0A Pending CN117392409A (en) 2022-07-01 2022-07-01 Work clothes identification method, device, equipment and computer readable storage medium

Country Status (1)

Country Link
CN (1) CN117392409A (en)

Similar Documents

Publication Publication Date Title
Spreeuwers et al. Towards robust evaluation of face morphing detection
CN111626371B (en) Image classification method, device, equipment and readable storage medium
CN107463865B (en) Face detection model training method, face detection method and device
CN110598019B (en) Repeated image identification method and device
CN113033465A (en) Living body detection model training method, device, equipment and storage medium
CN110866466A (en) Face recognition method, face recognition device, storage medium and server
CN111696080B (en) Face fraud detection method, system and storage medium based on static texture
CN112381104A (en) Image identification method and device, computer equipment and storage medium
CN111222433B (en) Automatic face auditing method, system, equipment and readable storage medium
CN110033388B (en) Social relationship establishing method and device and server
CN112036284B (en) Image processing method, device, equipment and storage medium
CN112651333A (en) Silence living body detection method and device, terminal equipment and storage medium
CN115223022B (en) Image processing method, device, storage medium and equipment
CN115116111B (en) Anti-disturbance human face living body detection model training method and device and electronic equipment
CN115273184B (en) Training method and device for human face living body detection model
CN117392409A (en) Work clothes identification method, device, equipment and computer readable storage medium
Mukherjee et al. Energy efficient face recognition in mobile-fog environment
CN114387651B (en) Face recognition method, device, equipment and storage medium
CN114038030A (en) Image tampering identification method, device and computer storage medium
CN115019052A (en) Image recognition method and device, electronic equipment and storage medium
CN117218398A (en) Data processing method and related device
CN116798087A (en) Employee on-duty state detection method and system
CN114140320A (en) Image migration method and training method and device of image migration model
CN114022699A (en) Image classification method and device, computer equipment and storage medium
CN113705307A (en) Image processing method, device, equipment and storage medium

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination