CN110689046A - Image recognition method, image recognition device, computer device, and storage medium - Google Patents
Image recognition method, image recognition device, computer device, and storage medium Download PDFInfo
- Publication number
- CN110689046A CN110689046A CN201910792041.1A CN201910792041A CN110689046A CN 110689046 A CN110689046 A CN 110689046A CN 201910792041 A CN201910792041 A CN 201910792041A CN 110689046 A CN110689046 A CN 110689046A
- Authority
- CN
- China
- Prior art keywords
- query
- similarity
- image
- reference images
- reference image
- 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
Links
- 238000000034 method Methods 0.000 title claims abstract description 43
- 238000013507 mapping Methods 0.000 claims abstract description 18
- 238000010606 normalization Methods 0.000 claims description 37
- 238000013528 artificial neural network Methods 0.000 claims description 23
- 238000012545 processing Methods 0.000 claims description 22
- 230000004913 activation Effects 0.000 claims description 9
- 238000004364 calculation method Methods 0.000 claims description 6
- 238000000605 extraction Methods 0.000 claims description 6
- 238000010276 construction Methods 0.000 claims description 3
- 238000003384 imaging method Methods 0.000 abstract 1
- 239000013074 reference sample Substances 0.000 description 28
- 230000006870 function Effects 0.000 description 14
- 239000000523 sample Substances 0.000 description 8
- 241000001667 Eueretagrotis sigmoides Species 0.000 description 4
- 238000004891 communication Methods 0.000 description 4
- 238000010586 diagram Methods 0.000 description 4
- 238000012886 linear function Methods 0.000 description 4
- 238000012544 monitoring process Methods 0.000 description 4
- 239000000284 extract Substances 0.000 description 3
- 238000004422 calculation algorithm Methods 0.000 description 2
- 230000004927 fusion Effects 0.000 description 2
- 239000000203 mixture Substances 0.000 description 2
- 238000004590 computer program Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 230000010365 information processing Effects 0.000 description 1
- 230000003993 interaction Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000012806 monitoring device Methods 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/22—Matching criteria, e.g. proximity measures
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
-
- 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
- G06N3/045—Combinations of networks
-
- 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
- G06N3/048—Activation functions
-
- 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
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Computing Systems (AREA)
- Software Systems (AREA)
- Molecular Biology (AREA)
- Computational Linguistics (AREA)
- Biophysics (AREA)
- Biomedical Technology (AREA)
- Mathematical Physics (AREA)
- General Health & Medical Sciences (AREA)
- Health & Medical Sciences (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
- Image Analysis (AREA)
Abstract
The invention provides an image recognition method, which comprises the following steps: acquiring a query image and a plurality of reference images; imaging the query image and each reference image, and extracting similarity characteristics of the image pairs; constructing a complete graph by taking the image pairs as nodes; calculating the similarity score of every two reference images, and determining the weight of the corresponding edge of every two reference images in the complete image according to the similarity score of every two reference images; mapping the similarity characteristics of the image pairs into message characteristics through a message network; updating the similarity characteristic of the image pair according to the message characteristic and the weight of the edge; calculating a similarity score of the image pair according to the updated similarity features; and determining a reference image matched with the query image according to the similarity score. The invention also provides an image recognition device, a computer device and a computer readable storage medium. The method and the device utilize the similarity information between the reference images to update the similarity characteristics of the query image and the reference images, and improve the accuracy of image identification.
Description
Technical Field
The present invention relates to the field of image recognition technologies, and in particular, to an image recognition method, an image recognition device, a computer device, and a computer-readable storage medium.
Background
At present, when a query image is matched with a reference image, only the similarity between the query image and the reference image is considered, and the similarity between the reference images is ignored. If the similarity between the query image and the reference image is not calculated well, the accuracy of image matching is affected.
Disclosure of Invention
In view of the foregoing, it is desirable to provide an image recognition method, an image recognition apparatus, a computer apparatus, and a computer-readable storage medium, which improve the accuracy of image recognition by performing image recognition using similarity information between reference images.
A first aspect of the present application provides an image recognition method, the method comprising:
acquiring a query image and a plurality of reference images;
forming a query-reference image pair by the query image and each reference image, and extracting similarity characteristics of each query-reference image pair;
constructing a complete graph by taking the query-reference image pairs as nodes, wherein each reference image corresponds to one node;
calculating similarity scores between every two reference images in the multiple reference images, and determining the weight values of edges corresponding to the two reference images in the complete image according to the similarity scores between the two reference images;
mapping the similarity features of each query-reference image pair into message features through a message network;
updating the similarity characteristic of each query-reference image pair according to the message characteristic of each query-reference image pair and the weight of each edge in the complete graph;
calculating a similarity score for each query-reference image pair based on the updated similarity features for each query-reference image pair;
determining images from the plurality of reference images that match the query image based on the similarity score for each query-reference image pair.
In another possible implementation, the extracting the similarity feature of each query-reference image pair includes:
inputting the query-reference image pair into a twin neural network based on a residual error network to obtain a feature map of the query image and a feature map of a reference image in the query-reference image pair;
subtracting the feature map of the query image from the feature map of the reference image to obtain a first feature map;
performing element-by-element squaring operation on the first feature map to obtain a second feature map;
and carrying out batch normalization processing on the second feature graph to obtain the similarity feature of the query-reference image pair.
In another possible implementation manner, the weights of the edges corresponding to the two reference images in the complete graph are:
wherein S (g)i,gj) Is the similarity of the reference images i, j.
In another possible implementation manner, the mapping the similarity feature of each query-reference image pair to a message feature through a message network includes:
classifying the similarity features of the query-reference image pair through a full connection layer of the message network to obtain the classified similarity features;
carrying out batch normalization processing on the classified similarity features through a batch normalization layer of the message network to obtain similarity features after normalization processing;
and converting the linear factors in the similarity features after the normalization processing into nonlinear factors through an activation layer of the message network to obtain the message features of the query-reference image pair.
In another possible implementation, the determining, from the plurality of reference images, an image that matches the query image according to the similarity score of each query-reference image pair includes:
determining a reference image with the highest similarity score in the plurality of reference images as an image matching the query image; or
And determining the reference images with similarity scores higher than a preset value in the plurality of reference images as the images matched with the query image.
In another possible implementation manner, the calculating the similarity score between each two reference images of the plurality of reference images includes:
performing region division on the two reference images according to the same division method;
calculating the logarithm relative RGB coordinates of each pixel point of each area of the two reference images;
clustering the pixel points in each region of the two reference images according to the logarithm relative RGB coordinates of each pixel point in each region of the two reference images to obtain a clustering center of each region of the two reference images;
determining a similarity score between the two reference images according to the clustering center of each region of the two reference images.
In another possible implementation manner, the calculating the similarity score between each two reference images of the plurality of reference images includes:
respectively inputting the two reference images into a first depth residual error network and a second depth residual error network, obtaining the overall characteristics of the two reference images from the first depth residual error network, and obtaining the local characteristics of the two reference images from the second depth residual error network;
and calculating a similarity score between the two reference images according to the overall characteristics and the local characteristics of the two reference images.
A second aspect of the present application provides an image recognition apparatus, the apparatus comprising:
an acquisition module for acquiring a query image and a plurality of reference images;
the extraction module is used for forming the query image and each reference image into a query-reference image pair and extracting the similarity characteristic of each query-reference image pair;
a construction module for constructing a complete graph with the query-reference image pairs as nodes, each reference image corresponding to a node;
a first determining module, configured to calculate a similarity score between each two reference images in the multiple reference images, and determine a weight of an edge corresponding to the two reference images in the complete graph according to the similarity score between the two reference images;
a mapping module for mapping the similarity features of each query-reference image pair to message features over a message network;
the updating module is used for updating the similarity characteristic of each query-reference image pair according to the message characteristic of each query-reference image pair and the weight of each edge in the complete graph;
a calculation module for calculating a similarity score for each query-reference image pair based on the updated similarity features for each query-reference image pair;
a second determination module to determine images from the plurality of reference images that match the query image based on the similarity score for each query-reference image pair.
In another possible implementation manner, the extracting, by the extraction module, the similarity feature of each query-reference image pair includes:
inputting the query-reference image pair into a twin neural network based on a residual error network to obtain a feature map of the query image and a feature map of a reference image in the query-reference image pair;
subtracting the feature map of the query image from the feature map of the reference image to obtain a first feature map;
performing element-by-element squaring operation on the first feature map to obtain a second feature map;
and carrying out batch normalization processing on the second feature graph to obtain the similarity feature of the query-reference image pair.
In another possible implementation manner, the weights of the edges corresponding to the two reference images in the complete graph are:
wherein S (g)i,gj) Is the similarity of the reference images i, j.
In another possible implementation manner, the mapping module is specifically configured to:
classifying the similarity features of the query-reference image pair through a full connection layer of the message network to obtain the classified similarity features;
carrying out batch normalization processing on the classified similarity features through a batch normalization layer of the message network to obtain similarity features after normalization processing;
and converting the linear factors in the similarity features after the normalization processing into nonlinear factors through an activation layer of the message network to obtain the message features of the query-reference image pair.
In another possible implementation manner, the second determining module is specifically configured to:
determining a reference image with the highest similarity score in the plurality of reference images as an image matching the query image; or
And determining the reference images with similarity scores higher than a preset value in the plurality of reference images as the images matched with the query image.
In another possible implementation manner, the calculation module is specifically configured to:
performing region division on the two reference images according to the same division method;
calculating the logarithm relative RGB coordinates of each pixel point of each area of the two reference images;
clustering the pixel points in each region of the two reference images according to the logarithm relative RGB coordinates of each pixel point in each region of the two reference images to obtain a clustering center of each region of the two reference images;
determining a similarity score between the two reference images according to the clustering center of each region of the two reference images.
In another possible implementation manner, the calculation module is specifically configured to:
respectively inputting the two reference images into a first depth residual error network and a second depth residual error network, obtaining the overall characteristics of the two reference images from the first depth residual error network, and obtaining the local characteristics of the two reference images from the second depth residual error network;
and calculating a similarity score between the two reference images according to the overall characteristics and the local characteristics of the two reference images.
A third aspect of the application provides a computer apparatus comprising a memory for storing at least one instruction and a processor for executing the at least one instruction to implement the image recognition method.
A fourth aspect of the present application provides a computer-readable storage medium having stored thereon at least one instruction which, when executed by a processor, implements the image recognition method.
It is seen from the above technical solutions that the present invention provides an image recognition method, apparatus, computer apparatus and computer readable storage medium, the method obtains a query image and a plurality of reference images; forming a query-reference image pair by the query image and each reference image, and extracting similarity characteristics of each query-reference image pair; constructing a complete graph by taking the query-reference image pairs as nodes, wherein each reference image corresponds to one node; calculating similarity scores between every two reference images in the multiple reference images, and determining the weight values of edges corresponding to the two reference images in the complete image according to the similarity scores between the two reference images; mapping the similarity features of each query-reference image pair into message features through a message network; updating the similarity characteristic of each query-reference image pair according to the message characteristic of each query-reference image pair and the weight of each edge in the complete graph; calculating a similarity score for each query-reference image pair based on the updated similarity features for each query-reference image pair; determining images from the plurality of reference images that match the query image based on the similarity score for each query-reference image pair. The method and the device utilize the similarity information between the reference images to update the similarity characteristics of the query-reference image pair, and improve the accuracy of image identification.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly described below, it is obvious that the drawings in the following description are only embodiments of the present invention, and for those skilled in the art, other drawings can be obtained according to the provided drawings without creative efforts.
FIG. 1 is a flow chart of an image recognition method provided by an embodiment of the invention;
FIG. 2 is a functional block diagram of an image recognition apparatus according to an embodiment of the present invention;
FIG. 3 is a schematic structural diagram of a computer device according to a preferred embodiment of the present invention.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
In order to make the aforementioned objects, features and advantages of the present invention comprehensible, embodiments accompanied with figures are described in further detail below.
Fig. 1 is a flowchart of an image recognition method according to an embodiment of the present invention. The image identification method is applied to a computer device and used for matching a query image with a reference image. The order of the steps in the flow chart may be changed and some steps may be omitted according to different needs.
S11, a query image and a plurality of reference images are obtained.
The query image is an image to be identified and the reference image is a known image. The method identifies an image from the plurality of reference images that contains the same content (e.g., an object or person) as the query image. For example, when object identification is required, the query image is an image containing an unknown object, the reference images are images containing known objects, and the method identifies a reference image containing the same object as the query image from the reference images. For another example, when person identification is required, the query image is an image containing an unknown person, the reference images are images containing known persons, and the method identifies an image containing the same person as the query image from the reference images.
The query image may be received from an external device. For example, a monitoring image captured by an external camera is acquired, and the monitoring image captured by the external camera is used as the query image.
Or, the computer device may include a camera, and may control a built-in camera of the computer device to capture an image, where the image captured by the built-in camera is used as the query image.
Alternatively, an image stored in advance by the computer device may be read, and the read image stored in advance may be used as the query image.
Alternatively, an image may be downloaded from a network, and the downloaded image may be used as the query image.
The plurality of reference images may be obtained from a preset image library. For example, when person identification is performed, the plurality of reference images may be acquired from a person image library.
And S12, forming a query-reference image pair by the query image and each reference image, and extracting the similarity characteristic of each query-reference image pair.
In this embodiment, the query image is paired with each reference image to form a plurality of query-reference image pairs.
A twin neural network based on a residual network may be utilized to extract similarity features for each query-reference image pair.
Preferably, said extracting similarity features for each query-reference image pair comprises:
inputting the query-reference image pair into a twin neural network based on a residual error network to obtain a feature map of the query image and a feature map of a reference image in the query-reference image pair;
subtracting the feature map of the query image from the feature map of the reference image to obtain a first feature map;
performing element-by-element squaring operation on the first feature map to obtain a second feature map;
and carrying out batch normalization processing on the second feature graph to obtain the similarity feature of the query-reference image pair.
The twin neural network based on the residual error network is two connected neural networks sharing a weight value, wherein one of the two connected neural networks takes the query image as input, and the other one takes a reference image in the query-reference image pair as input.
In this embodiment, the twin neural network is trained using a query-reference sample image pair in advance. The query-reference sample image pair is an image pair of the query sample image and the reference sample image. Each query-reference sample image pair has a label indicating whether the query sample image and the reference sample image in the query-reference sample image pair contain the same content. If the query sample image and the reference sample image contain the same content (e.g., the same person), the tag may be 1. The label may be 0 if the query sample image and the reference sample image contain different content (e.g., different people).
When the twin neural network is trained, extracting the similarity characteristics of the query-reference sample image pair, inputting the similarity characteristics of the query-reference sample image pair into a linear classifier to obtain the similarity score of the query-reference sample image pair, and obtaining the similarity score according to the similarity scoreCalculating a loss function with the labels of the query-reference sample image pair, and adjusting parameters of the twin neural network to minimize the loss function. Wherein the linear classifier may be a non-linear function, i.e. sigmoid function, of the formula f (x) 1/(1+ e)-x). The loss function may be:
wherein d isiIs a similarity feature of the ith query-reference sample image pair, F () represents a linear classifier, yiA label representing the ith query-reference sample image pair.
And S13, constructing a complete graph by taking the query-reference image pairs as nodes, wherein each reference image corresponds to one node.
The complete graph is a simple graph with an edge connected between each pair of nodes, that is, the nodes in the complete graph are connected two by two.
In this embodiment, each node of the full graph represents a query-reference image pair, and each edge of the full graph corresponds to two reference images and represents a relationship between the two reference images.
S14, calculating similarity scores between every two reference images in the multiple reference images, and determining the weight values of edges corresponding to the two reference images in the complete graph according to the similarity scores between the two reference images.
In this embodiment, the calculating the similarity score between each two reference images of the plurality of reference images includes:
performing region division on the two reference images according to the same division method;
calculating the logarithm relative RGB coordinates of each pixel point of each area of the two reference images;
clustering the pixel points in each region of the two reference images according to the logarithm relative RGB coordinates of each pixel point in each region of the two reference images to obtain a clustering center of each region of the two reference images;
determining a similarity score between the two reference images according to the clustering center of each region of the two reference images.
The two reference images may be each divided into upper and lower two regions or left and right two regions. It is also possible to divide the two reference pictures into more than two regions each, for example into three regions or four regions each.
The red component being RiGreen component is GiAnd the logarithm of the pixel point i with the blue component of Bi is (x) relative to the RGB coordinatei,yi) Which is
In (1),
can take the logarithm of the base e, i.e.
Alternatively, the base logarithm may be taken to be other values, such as base 10 logarithm.
The pixels in each region of the two reference images may be clustered using GMM (Gaussian Mixture Model) or K-Means algorithm to obtain a cluster center of each region of the two reference images.
The distance of the cluster centers of each of the two reference images may be calculated, and the similarity between the two reference images may be determined based on the distance of the cluster centers of each of the two reference images. The weighted sum of the distances of the cluster centers of each region of the two reference images may be taken as the similarity between the two reference images. The distance of the cluster centers of each region of the two reference images may be euclidean distance, manhattan distance, mahalanobis distance, or the like.
In another embodiment, the two reference images may be input into a neural network to extract features, and a similarity score between the two reference images may be calculated according to the features of the two reference images. For example, the two reference images are respectively input into a first depth residual error network and a second depth residual error network, the overall characteristics of the two reference images are obtained from the first depth residual error network, the local characteristics of the two reference images are obtained from the second depth residual error network, and the similarity score between the two reference images is calculated according to the overall characteristics and the local characteristics of the two reference images.
In this embodiment, the weight of the edge corresponding to the two reference images in the complete graph may be represented as:
wherein, S (g)i,gj) Is the similarity of the reference images i, j.
In another embodiment, each two reference images in the plurality of reference images may be combined into a reference image pair, the similarity feature of each reference image pair is extracted, and the similarity feature of each reference image pair is input into a linear classifier to obtain the similarity score of each reference image pair. Extracting similarity features for each reference image pair may be referred to S12.
And S15, mapping the similarity characteristic of each query-reference image pair into a message characteristic through a message network.
The messaging network is a neural network. In this embodiment, the message network is composed of a fully connected layer, a bulk normalization layer, and an active layer.
The similarity feature of the ith query-reference image pair is denoted as diThe message characteristic of the ith query-reference image pair is denoted as ti，ti＝F(di) I is 1,2, … N (indicating that there are N reference pictures).
Preferably, the mapping the similarity feature of each query-reference image pair to a message feature through a message network comprises:
the full connection layer of the message network classifies the similarity characteristics of the query-reference image pair to obtain the classified similarity characteristics;
the batch normalization layer of the message network performs batch normalization processing on the classified similarity features to obtain the similarity features after normalization processing;
and the activation layer of the message network converts the linear factors in the similarity characteristics after the normalization processing into nonlinear factors to obtain the message characteristics of the query-reference image pair.
In this embodiment, the fully-connected layer of the message network classifies the similarity features of the query-reference image pair, so that the similarity features with high similarity enter the next layer of the message network, that is, the batch normalization layer. The batch normalization layer in the message network continuously adjusts the intermediate output of the message network by using the mean value and the standard deviation on the small batch, so that the numerical value of the intermediate output of the whole message network in each layer is more stable. The activation layer in the message network converts the linear factors in the similarity characteristics transferred by the previous layer (namely the batch normalization layer) into the nonlinear factors through the activation function, so that the problem that the linear factors cannot be solved is solved.
In this embodiment, two message networks may be used to map the similarity features of each query-reference image pair to message features. For example, a message network consisting of a fully connected layer, a batch normalization layer, and an active layer is connected after a message network consisting of a fully connected layer, a batch normalization layer, and an active layer. More accurate depth characteristic information can be extracted through the two-layer message network.
And S16, updating the similarity feature of each query-reference image pair according to the message feature of each query-reference image pair and the weight value of each edge in the complete graph.
And updating the similarity feature of each query-reference image pair according to the message feature of each query-reference image pair and the weight value of each edge in the complete graph, namely updating the similarity feature of each query-reference image pair according to the connection relation between each node and other nodes of the complete graph. For each node in the complete graph, using the message characteristics of the query-reference image pair corresponding to other nodes connected with the node as the input characteristics of the node, and updating the similarity characteristics of the query-reference image pair corresponding to the node into weighted fusion of all the input characteristics and the original similarity characteristics, namely
Wherein
Indicating the updated ith similarity characteristic,
indicating the i-th similarity characteristic before updating,
representing the message characteristics from node j and alpha representing a weighting parameter that balances the fused characteristics and the original characteristics.
The similarity features of the query-reference image pair may be iteratively updated as follows:
and S17, calculating a similarity score of each query-reference image pair according to the updated similarity characteristics of each query-reference image pair.
In this embodiment, the updated similarity features of each query-reference image pair may be input into a linear classifier to obtain a similarity score of each query-reference image pair.
The linear classifier may be a non-linear function, i.e. sigmoid function, of the formula f (x) 1/(1+ e)-x)。
S18, determining images from the plurality of reference images that match the query image based on the similarity score for each query-reference image pair.
Preferably, the determining, from the plurality of reference images, an image that matches the query image according to the similarity score for each query-reference image pair comprises:
determining a reference image with the highest similarity score in the plurality of reference images as an image matching the query image; or
And determining the reference images with similarity scores higher than a preset value in the plurality of reference images as the images matched with the query image.
For example, there are 20 reference images, and the reference image having the highest similarity score among the 20 reference images is determined as the image matching the query image, or the reference image having a similarity score higher than 0.9 among the 20 reference images is determined as the image matching the query image.
The image identification method of the invention obtains a query image and a plurality of reference images; forming a query-reference image pair by the query image and each reference image, and extracting similarity characteristics of each query-reference image pair; constructing a complete graph by taking the query-reference image pairs as nodes, wherein each reference image corresponds to one node; calculating similarity scores between every two reference images in the multiple reference images, and determining the weight values of edges corresponding to the two reference images in the complete image according to the similarity scores between the two reference images; mapping the similarity features of each query-reference image pair into message features through a message network; updating the similarity characteristic of each query-reference image pair according to the message characteristic of each query-reference image pair and the weight of each edge in the complete graph; calculating a similarity score for each query-reference image pair based on the updated similarity features for each query-reference image pair; determining images from the plurality of reference images that match the query image based on the similarity score for each query-reference image pair. The method utilizes the similarity information between the reference images to update the similarity characteristics of the query-reference image pair, and improves the accuracy of image identification.
As shown in fig. 2, fig. 2 is a functional block diagram of an image recognition apparatus according to an embodiment of the present invention. The image recognition device 20 includes an acquisition module 210, an extraction module 220, a construction module 230, a first determination module 240, a mapping module 250, an update module 260, a calculation module 270, and a second determination module 280. The module referred to in the present invention refers to a series of computer program segments capable of being executed by a processor of a computer device and performing a fixed function, which are stored in a memory of the computer device.
The obtaining module 210 is configured to obtain a query image and a plurality of reference images.
The query image is an image to be identified and the reference image is a known image. The method identifies an image from the plurality of reference images that contains the same content (e.g., an object or person) as the query image. For example, when object identification is required, the query image is an image containing an unknown object, the reference images are images containing known objects, and the method identifies a reference image containing the same object as the query image from the reference images. For another example, when person identification is required, the query image is an image containing an unknown person, the reference images are images containing known persons, and the method identifies an image containing the same person as the query image from the reference images.
The query image may be received from an external device. For example, a monitoring image captured by an external camera is acquired, and the monitoring image captured by the external camera is used as the query image.
Or, the computer device may include a camera, and may control a built-in camera of the computer device to capture an image, where the image captured by the built-in camera is used as the query image.
Alternatively, an image stored in advance by the computer device may be read, and the read image stored in advance may be used as the query image.
Alternatively, an image may be downloaded from a network, and the downloaded image may be used as the query image.
The plurality of reference images may be obtained from a preset image library. For example, when person identification is performed, the plurality of reference images may be acquired from a person image library.
The extracting module 220 is configured to combine the query image and each reference image into a query-reference image pair, and extract a similarity feature of each query-reference image pair.
In this embodiment, the query image is paired with each reference image to form a plurality of query-reference image pairs.
A twin neural network based on a residual network may be utilized to extract similarity features for each query-reference image pair.
Preferably, the extraction module extracts similarity features of each query-reference image pair, in particular for:
inputting the query-reference image pair into a twin neural network based on a residual error network to obtain a feature map of the query image and a feature map of a reference image in the query-reference image pair;
subtracting the feature map of the query image from the feature map of the reference image to obtain a first feature map;
performing element-by-element squaring operation on the first feature map to obtain a second feature map;
and carrying out batch normalization processing on the second feature graph to obtain the similarity feature of the query-reference image pair.
The twin neural network based on the residual error network is two connected neural networks sharing a weight value, wherein one of the two connected neural networks takes the query image as input, and the other one takes a reference image in the query-reference image pair as input.
In this embodiment, the twin neural network is trained using a query-reference sample image pair in advance. The query-reference sample image pair is an image pair of the query sample image and the reference sample image. Each query-reference sample image pair has a label indicating whether the query sample image and the reference sample image in the query-reference sample image pair contain the same content. If the query sample image and the reference sample image contain the same content (e.g., the same person), the tag may be 1. The label may be 0 if the query sample image and the reference sample image contain different content (e.g., different people).
When the twin neural network is trained, extracting the similarity characteristics of the query-reference sample image pair, inputting the similarity characteristics of the query-reference sample image pair into a linear classifier to obtain the similarity score of the query-reference sample image pair, calculating a loss function according to the similarity score and the label of the query-reference sample image pair, and adjusting the parameters of the twin neural network to minimize the loss function. Wherein the linear classifier may be a non-linear function, i.e. sigmoid function, of the formula f (x) 1/(1+ e)-x). The loss function may be:
wherein d isiIs a similarity feature of the ith query-reference sample image pair, F () represents a linear classifier, yiA label representing the ith query-reference sample image pair.
The constructing module 230 is configured to construct a complete graph with the query-reference image pairs as nodes, where each reference image corresponds to a node.
The complete graph is a simple graph with an edge connected between each pair of nodes, that is, the nodes in the complete graph are connected two by two.
In this embodiment, each node of the full graph represents a query-reference image pair, and each edge of the full graph corresponds to two reference images and represents a relationship between the two reference images.
The first determining module 240 is configured to calculate a similarity score between each two reference images in the plurality of reference images, and determine a weight of an edge corresponding to the two reference images in the complete graph according to the similarity score between the two reference images.
In this embodiment, the calculating module calculates a similarity score between each two reference images in the plurality of reference images, and is specifically configured to:
performing region division on the two reference images according to the same division method;
calculating the logarithm relative RGB coordinates of each pixel point of each area of the two reference images;
clustering the pixel points in each region of the two reference images according to the logarithm relative RGB coordinates of each pixel point in each region of the two reference images to obtain a clustering center of each region of the two reference images;
determining a similarity score between the two reference images according to the clustering center of each region of the two reference images.
The two reference images may be each divided into upper and lower two regions or left and right two regions. It is also possible to divide the two reference pictures into more than two regions each, for example into three regions or four regions each.
The red component being RiGreen component is GiAnd the logarithm of the pixel point i with the blue component of Bi is (x) relative to the RGB coordinatei,yi) Which is
In (1),can take the logarithm of the base e, i.e.
Alternatively, the base logarithm may be taken to be other values, such as base 10 logarithm.
The pixels in each region of the two reference images may be clustered using GMM (Gaussian Mixture Model) or K-Means algorithm to obtain a cluster center of each region of the two reference images.
The distance of the cluster centers of each of the two reference images may be calculated, and the similarity between the two reference images may be determined based on the distance of the cluster centers of each of the two reference images. The weighted sum of the distances of the cluster centers of each region of the two reference images may be taken as the similarity between the two reference images. The distance of the cluster centers of each region of the two reference images may be euclidean distance, manhattan distance, mahalanobis distance, or the like.
In another embodiment, the two reference images may be input into a neural network to extract features, and a similarity score between the two reference images may be calculated according to the features of the two reference images. For example, the two reference images are respectively input into a first depth residual error network and a second depth residual error network, the overall characteristics of the two reference images are obtained from the first depth residual error network, the local characteristics of the two reference images are obtained from the second depth residual error network, and the similarity score between the two reference images is calculated according to the overall characteristics and the local characteristics of the two reference images.
In this embodiment, the weight of the edge corresponding to the two reference images in the complete graph may be represented as:
wherein, S (g)i,gj) Is the similarity of the reference images i, j.
In another embodiment, each two reference images in the plurality of reference images may be combined into a reference image pair, the similarity feature of each reference image pair is extracted, and the similarity feature of each reference image pair is input into a linear classifier to obtain the similarity score of each reference image pair. Extracting the similarity features for each reference image pair may extract the similarity features for each query-reference image pair with reference to the extraction module 220.
The mapping module 250 is configured to map the similarity features of each query-reference image pair into message features through a message network.
The messaging network is a neural network. In this embodiment, the message network is composed of a fully connected layer, a bulk normalization layer, and an active layer.
The similarity feature of the ith query-reference image pair is denoted as diThe message characteristic of the ith query-reference image pair is denoted as ti，ti＝F(di) I is 1,2, … N (indicating that there are N reference pictures).
Preferably, the mapping module maps the similarity feature of each query-reference image pair to a message feature through a message network, and is specifically configured to:
the full connection layer of the message network classifies the similarity characteristics of the query-reference image pair to obtain the classified similarity characteristics;
the batch normalization layer of the message network performs batch normalization processing on the classified similarity features to obtain the similarity features after normalization processing;
and the activation layer of the message network converts the linear factors in the similarity characteristics after the normalization processing into nonlinear factors to obtain the message characteristics of the query-reference image pair.
In this embodiment, the fully-connected layer of the message network classifies the similarity features of the query-reference image pair, so that the similarity features with high similarity enter the next layer of the message network, that is, the batch normalization layer. The batch normalization layer in the message network continuously adjusts the intermediate output of the message network by using the mean value and the standard deviation on the small batch, so that the numerical value of the intermediate output of the whole message network in each layer is more stable. The activation layer in the message network converts the linear factors in the similarity characteristics transferred by the previous layer (namely the batch normalization layer) into the nonlinear factors through the activation function, so that the problem that the linear factors cannot be solved is solved.
In this embodiment, two message networks may be used to map the similarity features of each query-reference image pair to message features. For example, a message network consisting of a fully connected layer, a batch normalization layer, and an active layer is connected after a message network consisting of a fully connected layer, a batch normalization layer, and an active layer. More accurate depth characteristic information can be extracted through the two-layer message network.
The updating module 260 is configured to update the similarity feature of each query-reference image pair according to the message feature of each query-reference image pair and the weight of each edge in the complete graph.
And updating the similarity feature of each query-reference image pair according to the message feature of each query-reference image pair and the weight value of each edge in the complete graph, namely updating the similarity feature of each query-reference image pair according to the connection relation between each node and other nodes of the complete graph. For each node in the complete graph, using the message characteristics of the query-reference image pair corresponding to other nodes connected with the node as the input characteristics of the node, and updating the similarity characteristics of the query-reference image pair corresponding to the node into weighted fusion of all the input characteristics and the original similarity characteristics, namely
Wherein
Indicating the updated ith similarity characteristic,
indicating the i-th similarity characteristic before updating,
representing the message characteristics from node j and alpha representing a weighting parameter that balances the fused characteristics and the original characteristics.
The similarity features of the query-reference image pair may be iteratively updated as follows:
the calculating module 270 is configured to calculate a similarity score for each query-reference image pair according to the updated similarity features of each query-reference image pair.
In this embodiment, the updated similarity features of each query-reference image pair may be input into a linear classifier to obtain a similarity score of each query-reference image pair.
The linear classifier may be a non-linear function, i.e. sigmoid function, of the formula f (x) 1/(1+ e)-x)。
The second determining module 280 is configured to determine images matching the query image from the plurality of reference images according to the similarity score of each query-reference image pair.
Preferably, the second determination module determines, from the plurality of reference images, an image matching the query image according to the similarity score of each query-reference image pair, in particular for:
determining a reference image with the highest similarity score in the plurality of reference images as an image matching the query image; or
And determining the reference images with similarity scores higher than a preset value in the plurality of reference images as the images matched with the query image.
For example, there are 20 reference images, and the reference image having the highest similarity score among the 20 reference images is determined as the image matching the query image, or the reference image having a similarity score higher than 0.9 among the 20 reference images is determined as the image matching the query image.
The image recognition apparatus 20 of the present invention acquires a query image and a plurality of reference images; forming a query-reference image pair by the query image and each reference image, and extracting similarity characteristics of each query-reference image pair; constructing a complete graph by taking the query-reference image pairs as nodes, wherein each reference image corresponds to one node; calculating similarity scores between every two reference images in the multiple reference images, and determining the weight values of edges corresponding to the two reference images in the complete image according to the similarity scores between the two reference images; mapping the similarity features of each query-reference image pair into message features through a message network; updating the similarity characteristic of each query-reference image pair according to the message characteristic of each query-reference image pair and the weight of each edge in the complete graph; calculating a similarity score for each query-reference image pair based on the updated similarity features for each query-reference image pair; determining images from the plurality of reference images that match the query image based on the similarity score for each query-reference image pair. The image recognition device 20 of the present invention updates the similarity characteristics of the query-reference image pair using the similarity information between the reference images, improving the accuracy of image recognition.
The integrated unit implemented in the form of a software functional module may be stored in a computer-readable storage medium. The software functional module is stored in a computer readable storage medium and includes several instructions for causing a computer device (which may be a personal computer, a server, or a network device) or a processor (processor) to execute some steps of the methods according to the embodiments of the present invention.
Fig. 3 is a schematic structural diagram of a computer device 3 according to a preferred embodiment of the present invention for implementing an image recognition display method. In the present embodiment, the computer device 3 comprises at least one transmitting device 31, at least one memory 32, at least one processor 33, at least one receiving device 34 and at least one communication bus. Wherein the communication bus is used for realizing connection communication among the components.
The computer device 3 is a device capable of automatically performing numerical calculation and/or information processing according to a preset or stored instruction, and its hardware includes, but is not limited to, a microprocessor, an Application Specific Integrated Circuit (ASIC), a Programmable Gate Array (FPGA), a Digital Signal Processor (DSP), an embedded device, and the like. The computer arrangement 3 may also comprise network equipment and/or user equipment. Wherein the network device includes, but is not limited to, a single network server, a server group consisting of a plurality of network servers, or a Cloud Computing (Cloud Computing) based Cloud consisting of a large number of hosts or network servers, wherein Cloud Computing is one of distributed Computing, a super virtual computer consisting of a collection of loosely coupled computers.
The computer device 3 may be, but is not limited to, any electronic product that can perform human-computer interaction with a user through a keyboard, a touch pad, or a voice control device, for example, a tablet computer, a smart phone, a monitoring device, and other terminals.
The Network where the computer device 3 is located includes, but is not limited to, the internet, a wide area Network, a metropolitan area Network, a local area Network, a Virtual Private Network (VPN), and the like.
The receiving device 34 and the transmitting device 31 may be wired transmitting ports, or may be wireless devices, for example, including antenna devices, for performing data communication with other devices.
The memory 32 is used to store program code. The memory 32 may be a memory bank, a TF Card (Trans-flash Card), a smart media Card (smart media Card), a secure digital Card (secure digital Card), a flash memory Card (flash Card), or other storage devices.
The processor 33 may comprise one or more microprocessors, digital processors. The processor 33 may call program code stored in the memory 32 to perform the associated functions. For example, the modules described in fig. 2 are program codes stored in the memory 32 and executed by the processor 33 to implement an image recognition display method. The processor 33 is also called a Central Processing Unit (CPU), and is an ultra-large scale integrated circuit, which is an operation Core (Core) and a Control Core (Control Unit).
In the embodiments provided in the present invention, it should be understood that the disclosed system, apparatus and method may be implemented in other ways. For example, the above-described apparatus embodiments are merely illustrative, and for example, the division of the modules is only one logical functional division, and other divisions may be realized in practice.
The modules described as separate parts may or may not be physically separate, and parts displayed as modules may or may not be physical units, may be located in one place, or may be distributed on a plurality of network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of the present embodiment.
In addition, functional modules in the embodiments of the present invention may be integrated into one processing unit, or each unit may exist alone physically, or two or more units are integrated into one unit. The integrated unit can be realized in a form of hardware, or in a form of hardware plus a software functional module.
It will be evident to those skilled in the art that the invention is not limited to the details of the foregoing illustrative embodiments, and that the present invention may be embodied in other specific forms without departing from the spirit or essential attributes thereof. The present embodiments are therefore to be considered in all respects as illustrative and not restrictive, the scope of the invention being indicated by the appended claims rather than by the foregoing description, and all changes which come within the meaning and range of equivalency of the claims are therefore intended to be embraced therein. Any reference signs in the claims shall not be construed as limiting the claim concerned. Furthermore, it is obvious that the word "comprising" does not exclude other elements or steps, and the singular does not exclude the plural. A plurality of units or means recited in the system claims may also be implemented by one unit or means in software or hardware. The terms second, etc. are used to denote names, but not any particular order.
Finally, it should be noted that the above embodiments are only for illustrating the technical solutions of the present invention and not for limiting, and although the present invention is described in detail with reference to the preferred embodiments, it should be understood by those skilled in the art that modifications or equivalent substitutions may be made on the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention.
Claims (10)
1. An image recognition method, characterized in that the method comprises:
acquiring a query image and a plurality of reference images;
forming a query-reference image pair by the query image and each reference image, and extracting similarity characteristics of each query-reference image pair;
constructing a complete graph by taking the query-reference image pairs as nodes, wherein each reference image corresponds to one node;
calculating similarity scores between every two reference images in the multiple reference images, and determining the weight values of edges corresponding to the two reference images in the complete image according to the similarity scores between the two reference images;
mapping the similarity features of each query-reference image pair into message features through a message network;
updating the similarity characteristic of each query-reference image pair according to the message characteristic of each query-reference image pair and the weight of each edge in the complete graph;
calculating a similarity score for each query-reference image pair based on the updated similarity features for each query-reference image pair;
determining images from the plurality of reference images that match the query image based on the similarity score for each query-reference image pair.
2. The method of claim 1, wherein said extracting similarity features for each query-reference image pair comprises:
inputting the query-reference image pair into a twin neural network based on a residual error network to obtain a feature map of the query image and a feature map of a reference image in the query-reference image pair;
subtracting the feature map of the query image from the feature map of the reference image to obtain a first feature map;
performing element-by-element squaring operation on the first feature map to obtain a second feature map;
and carrying out batch normalization processing on the second feature graph to obtain the similarity feature of the query-reference image pair.
3. The method of claim 1, wherein the weights of the edges corresponding to the two reference images in the complete graph are:
wherein S (g)i,gj) Is the similarity of the reference images i, j.
4. The method of claim 1, wherein mapping the similarity features of each query-reference image pair to message features over a message network comprises:
classifying the similarity features of the query-reference image pair through a full connection layer of the message network to obtain the classified similarity features;
carrying out batch normalization processing on the classified similarity features through a batch normalization layer of the message network to obtain similarity features after normalization processing;
and converting the linear factors in the similarity features after the normalization processing into nonlinear factors through an activation layer of the message network to obtain the message features of the query-reference image pair.
5. The method of claim 1, wherein determining images from the plurality of reference images that match the query image based on the similarity score for each query-reference image pair comprises:
determining a reference image with the highest similarity score in the plurality of reference images as an image matching the query image; or
And determining the reference images with similarity scores higher than a preset value in the plurality of reference images as the images matched with the query image.
6. The method of any one of claims 1-5, wherein the calculating the similarity score between each two of the plurality of reference images comprises:
performing region division on the two reference images according to the same division method;
calculating the logarithm relative RGB coordinates of each pixel point of each area of the two reference images;
clustering the pixel points in each region of the two reference images according to the logarithm relative RGB coordinates of each pixel point in each region of the two reference images to obtain a clustering center of each region of the two reference images;
determining a similarity score between the two reference images according to the clustering center of each region of the two reference images.
7. The method of any one of claims 1-5, wherein the calculating the similarity score between each two of the plurality of reference images comprises:
respectively inputting the two reference images into a first depth residual error network and a second depth residual error network, obtaining the overall characteristics of the two reference images from the first depth residual error network, and obtaining the local characteristics of the two reference images from the second depth residual error network;
and calculating a similarity score between the two reference images according to the overall characteristics and the local characteristics of the two reference images.
8. An image recognition apparatus, characterized in that the apparatus comprises:
an acquisition module for acquiring a query image and a plurality of reference images;
the extraction module is used for forming the query image and each reference image into a query-reference image pair and extracting the similarity characteristic of each query-reference image pair;
a construction module for constructing a complete graph with the query-reference image pairs as nodes, each reference image corresponding to a node;
a first determining module, configured to calculate a similarity score between each two reference images in the multiple reference images, and determine a weight of an edge corresponding to the two reference images in the complete graph according to the similarity score between the two reference images;
a mapping module for mapping the similarity features of each query-reference image pair to message features over a message network;
the updating module is used for updating the similarity characteristic of each query-reference image pair according to the message characteristic of each query-reference image pair and the weight of each edge in the complete graph;
a calculation module for calculating a similarity score for each query-reference image pair based on the updated similarity features for each query-reference image pair;
a second determination module to determine images from the plurality of reference images that match the query image based on the similarity score for each query-reference image pair.
9. A computer device, characterized in that the computer device comprises a memory for storing at least one instruction and a processor for executing the at least one instruction to implement the image recognition method according to any one of claims 1 to 7.
10. A computer-readable storage medium having stored thereon computer instructions, which when executed by a processor, implement the image recognition method of any one of claims 1 to 7.
Priority Applications (2)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910792041.1A CN110689046A (en) | 2019-08-26 | 2019-08-26 | Image recognition method, image recognition device, computer device, and storage medium |
| PCT/CN2020/086768 WO2021036309A1 (en) | 2019-08-26 | 2020-04-24 | Image recognition method and apparatus, computer apparatus, and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910792041.1A CN110689046A (en) | 2019-08-26 | 2019-08-26 | Image recognition method, image recognition device, computer device, and storage medium |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN110689046A true CN110689046A (en) | 2020-01-14 |
Family
ID=69108626
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910792041.1A Pending CN110689046A (en) | 2019-08-26 | 2019-08-26 | Image recognition method, image recognition device, computer device, and storage medium |
Country Status (2)
| Country | Link |
|---|---|
| CN (1) | CN110689046A (en) |
| WO (1) | WO2021036309A1 (en) |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111950493A (en) * | 2020-08-20 | 2020-11-17 | 华北电力大学 | Image recognition method and device, terminal equipment and readable storage medium |
| CN112307934A (en) * | 2020-10-27 | 2021-02-02 | 深圳市商汤科技有限公司 | Image detection method, and training method, device, equipment and medium of related model |
| WO2021036309A1 (en) * | 2019-08-26 | 2021-03-04 | 深圳壹账通智能科技有限公司 | Image recognition method and apparatus, computer apparatus, and storage medium |
| WO2024061311A1 (en) * | 2022-09-21 | 2024-03-28 | 北京沃东天骏信息技术有限公司 | Model training method and apparatus, and image classification method and apparatus |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116188805B (en) * | 2023-04-26 | 2023-08-04 | 青岛尘元科技信息有限公司 | Image content analysis method and device for massive images and image information network |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2011021605A1 (en) * | 2009-08-18 | 2011-02-24 | 公立大学法人大阪府立大学 | Object detection method |
| CN107871143A (en) * | 2017-11-15 | 2018-04-03 | 深圳云天励飞技术有限公司 | Image-recognizing method and device, computer installation and computer-readable recording medium |
| CN107895021A (en) * | 2017-11-15 | 2018-04-10 | 深圳云天励飞技术有限公司 | Image-recognizing method and device, computer installation and computer-readable recording medium |
| CN108985190A (en) * | 2018-06-28 | 2018-12-11 | 北京市商汤科技开发有限公司 | Target identification method and device, electronic equipment, storage medium, program product |
Family Cites Families (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| IL236598A0 (en) * | 2015-01-05 | 2015-05-31 | Superfish Ltd | Image similarity as a function of weighted descriptor similarities derived from neural networks |
| CN110689046A (en) * | 2019-08-26 | 2020-01-14 | 深圳壹账通智能科技有限公司 | Image recognition method, image recognition device, computer device, and storage medium |
-
2019
- 2019-08-26 CN CN201910792041.1A patent/CN110689046A/en active Pending
-
2020
- 2020-04-24 WO PCT/CN2020/086768 patent/WO2021036309A1/en active Application Filing
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2011021605A1 (en) * | 2009-08-18 | 2011-02-24 | 公立大学法人大阪府立大学 | Object detection method |
| CN107871143A (en) * | 2017-11-15 | 2018-04-03 | 深圳云天励飞技术有限公司 | Image-recognizing method and device, computer installation and computer-readable recording medium |
| CN107895021A (en) * | 2017-11-15 | 2018-04-10 | 深圳云天励飞技术有限公司 | Image-recognizing method and device, computer installation and computer-readable recording medium |
| CN108985190A (en) * | 2018-06-28 | 2018-12-11 | 北京市商汤科技开发有限公司 | Target identification method and device, electronic equipment, storage medium, program product |
Non-Patent Citations (1)
| Title |
|---|
| 庄小芳;滕忠坚;洪亲;: "基于图像混合图结构的区域匹配和图像约化算法", 计算机与现代化, no. 02 * |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2021036309A1 (en) * | 2019-08-26 | 2021-03-04 | 深圳壹账通智能科技有限公司 | Image recognition method and apparatus, computer apparatus, and storage medium |
| CN111950493A (en) * | 2020-08-20 | 2020-11-17 | 华北电力大学 | Image recognition method and device, terminal equipment and readable storage medium |
| CN111950493B (en) * | 2020-08-20 | 2024-03-08 | 华北电力大学 | Image recognition method, device, terminal equipment and readable storage medium |
| CN112307934A (en) * | 2020-10-27 | 2021-02-02 | 深圳市商汤科技有限公司 | Image detection method, and training method, device, equipment and medium of related model |
| WO2022088411A1 (en) * | 2020-10-27 | 2022-05-05 | 深圳市商汤科技有限公司 | Image detection method and apparatus, related model training method and apparatus, and device, medium and program |
| WO2024061311A1 (en) * | 2022-09-21 | 2024-03-28 | 北京沃东天骏信息技术有限公司 | Model training method and apparatus, and image classification method and apparatus |
Also Published As
| Publication number | Publication date |
|---|---|
| WO2021036309A1 (en) | 2021-03-04 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110662484B (en) | System and method for whole body measurement extraction | |
| CN110689046A (en) | Image recognition method, image recognition device, computer device, and storage medium | |
| CN110363091B (en) | Face recognition method, device and equipment under side face condition and storage medium | |
| WO2021082635A1 (en) | Region of interest detection method and apparatus, readable storage medium and terminal device | |
| WO2019119505A1 (en) | Face recognition method and device, computer device and storage medium | |
| EP4394726A1 (en) | Image classification method and apparatus, computer device, and storage medium | |
| CN111814620A (en) | Face image quality evaluation model establishing method, optimization method, medium and device | |
| CN113222942A (en) | Training method of multi-label classification model and method for predicting labels | |
| CN112598643A (en) | Depth counterfeit image detection and model training method, device, equipment and medium | |
| EP4113376A1 (en) | Image classification model training method and apparatus, computer device, and storage medium | |
| CN113490947A (en) | Detection model training method and device, detection model using method and storage medium | |
| CN116580257A (en) | Feature fusion model training and sample retrieval method and device and computer equipment | |
| CN111291887A (en) | Neural network training method, image recognition method, device and electronic equipment | |
| CN110765882A (en) | Video tag determination method, device, server and storage medium | |
| CN112036284B (en) | Image processing method, device, equipment and storage medium | |
| CN110489659A (en) | Data matching method and device | |
| CN111694954A (en) | Image classification method and device and electronic equipment | |
| CN113128526B (en) | Image recognition method and device, electronic equipment and computer-readable storage medium | |
| CN111340213A (en) | Neural network training method, electronic device, and storage medium | |
| CN114332993A (en) | Face recognition method and device, electronic equipment and computer readable storage medium | |
| CN112329666A (en) | Face recognition method and device, electronic equipment and storage medium | |
| CN111709473A (en) | Object feature clustering method and device | |
| CN112232890B (en) | Data processing method, device, equipment and storage medium | |
| CN113780148A (en) | Traffic sign image recognition model training method and traffic sign image recognition method | |
| CN111275183A (en) | Visual task processing method and device and electronic system |
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 |