CN111488945A - Image processing method, image processing device, computer equipment and computer readable storage medium - Google Patents
Image processing method, image processing device, computer equipment and computer readable storage medium Download PDFInfo
- Publication number
- CN111488945A CN111488945A CN202010306531.9A CN202010306531A CN111488945A CN 111488945 A CN111488945 A CN 111488945A CN 202010306531 A CN202010306531 A CN 202010306531A CN 111488945 A CN111488945 A CN 111488945A
- Authority
- CN
- China
- Prior art keywords
- image
- channel
- vehicle
- vehicle image
- sample
- 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
- 238000012545 processing Methods 0.000 title claims abstract description 36
- 238000003672 processing method Methods 0.000 title claims abstract description 25
- 238000013145 classification model Methods 0.000 claims abstract description 86
- 230000000875 corresponding effect Effects 0.000 claims abstract description 72
- 238000000034 method Methods 0.000 claims abstract description 60
- 238000010586 diagram Methods 0.000 claims abstract description 56
- 230000002596 correlated effect Effects 0.000 claims abstract description 14
- 238000012549 training Methods 0.000 claims description 32
- 238000009499 grossing Methods 0.000 claims description 24
- 230000008569 process Effects 0.000 claims description 24
- 238000004590 computer program Methods 0.000 claims description 23
- 230000006870 function Effects 0.000 claims description 20
- 230000003321 amplification Effects 0.000 claims description 8
- 238000003199 nucleic acid amplification method Methods 0.000 claims description 8
- 238000005516 engineering process Methods 0.000 description 10
- 208000037170 Delayed Emergence from Anesthesia Diseases 0.000 description 6
- 238000001514 detection method Methods 0.000 description 4
- 238000011176 pooling Methods 0.000 description 4
- 238000000605 extraction Methods 0.000 description 3
- 230000002159 abnormal effect Effects 0.000 description 2
- 230000006835 compression Effects 0.000 description 2
- 238000007906 compression Methods 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 238000007689 inspection Methods 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000013473 artificial intelligence Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000004891 communication Methods 0.000 description 1
- 238000013135 deep learning Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000009977 dual effect Effects 0.000 description 1
- 230000007613 environmental effect Effects 0.000 description 1
- 238000010191 image analysis Methods 0.000 description 1
- 230000003993 interaction Effects 0.000 description 1
- 238000007781 pre-processing Methods 0.000 description 1
- 230000003068 static effect Effects 0.000 description 1
- 230000001360 synchronised effect Effects 0.000 description 1
- 238000012795 verification Methods 0.000 description 1
- 230000000007 visual effect Effects 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
- G06F18/25—Fusion techniques
- G06F18/253—Fusion techniques of extracted features
-
- 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
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V2201/00—Indexing scheme relating to image or video recognition or understanding
- G06V2201/08—Detecting or categorising vehicles
Abstract
The application relates to an image processing method, an image processing device, a computer device and a computer readable storage medium. The image processing method comprises the following steps: acquiring a vehicle image to be classified; acquiring a multi-channel characteristic diagram corresponding to the vehicle image and a weight coefficient of each channel by adopting an image classification model; the weight coefficient of each channel is positively correlated with the characteristic importance of the corresponding channel; acquiring a target characteristic diagram of the vehicle image according to the multi-channel characteristic diagram and the weight coefficient of each channel; and acquiring the image category of the vehicle image based on the target feature map. By adopting the method, the image classification accuracy of the vehicle image by the computer equipment can be improved.
Description
Technical Field
The present invention relates to the field of image processing technologies, and in particular, to an image processing method, an image processing apparatus, a computer device, and a computer-readable storage medium.
Background
With the rapid development of artificial intelligence technology, image classification is used as a basic technology in the field of computer vision, and is also increasingly applied to multiple fields, such as communication, traffic, medical image analysis, aerospace and other fields.
Taking the traffic field as an example, the quantity of motor vehicles in China is continuously increased in recent years, the image classification technology is applied to the vehicle detection process, and a deep learning classification model is used for replacing manpower to classify dynamic identifiers of vehicle chassis in different shapes (rectangle, square or circle), judge the on-off state of vehicle lamps, judge whether a truck box is abnormal or not and the like, so that the labor cost can be greatly saved, and the vehicle detection efficiency is improved.
However, the image classification method has a problem of low classification accuracy.
Disclosure of Invention
In view of the foregoing, it is desirable to provide an image processing method, an apparatus, a computer device, and a computer-readable storage medium capable of improving the classification accuracy of vehicle images in view of the above technical problems.
In a first aspect, an embodiment of the present application provides an image processing method, where the image processing method includes:
acquiring a vehicle image to be classified;
acquiring a multi-channel characteristic diagram corresponding to the vehicle image and a weight coefficient of each channel by adopting an image classification model; the weight coefficient of each channel is positively correlated with the characteristic importance of the corresponding channel;
acquiring a target characteristic diagram of the vehicle image according to the multi-channel characteristic diagram and the weight coefficient of each channel;
and acquiring the image category of the vehicle image based on the target feature map.
In one embodiment, the obtaining the target feature map of the vehicle image according to the multi-channel feature map and the weight coefficient of each channel includes:
multiplying each channel of the multichannel feature map by a corresponding weight coefficient respectively to obtain a weighted multichannel feature map;
and correspondingly adding the pixel value of each pixel point of the vehicle image and the pixel value of each pixel point of the multichannel characteristic diagram after the weight adjustment to obtain the target characteristic diagram.
In one embodiment, the image classification model is a residual network model.
In one embodiment, after acquiring the vehicle image to be classified, the method further includes:
according to the required input size of the image classification model, the size of the vehicle image is adjusted by adopting a nearest neighbor interpolation method to obtain an adjusted vehicle image;
correspondingly, the obtaining of the multi-channel feature map corresponding to the vehicle image and the weight coefficient of each channel by using the image classification model includes:
and acquiring a multi-channel characteristic diagram corresponding to the adjusted vehicle image and a weight coefficient of each channel by adopting the image classification model.
In one embodiment, the training process of the image classification model includes:
obtaining a plurality of sample vehicle images;
and replacing the loss function of the initial residual error network model with focal loss, and training the replaced initial residual error network model according to the plurality of sample vehicle images to obtain the image classification model.
In one embodiment, the training the replaced initial residual error network model according to the plurality of sample vehicle images to obtain the image classification model includes:
inputting the plurality of sample vehicle images into the replaced initial residual error network model to obtain the output category of each sample vehicle image;
performing label smoothing processing on the output category of each sample vehicle image according to the output category of each sample vehicle image;
and adjusting the parameters of the initial residual error network model according to the output category after the label smoothing processing corresponding to each sample vehicle image.
In one embodiment, the obtaining a plurality of sample vehicle images comprises:
acquiring a plurality of original sample images;
performing data amplification processing on the plurality of original sample images to obtain a plurality of amplified sample images;
and adjusting the sizes of the plurality of amplified sample images by adopting a nearest neighbor interpolation method to obtain the plurality of sample vehicle images.
In a second aspect, an embodiment of the present application provides an image processing apparatus, including:
the image acquisition module is used for acquiring a vehicle image to be classified;
the first characteristic acquisition module is used for acquiring a multi-channel characteristic diagram corresponding to the vehicle image and a weight coefficient of each channel by adopting an image classification model; the weight coefficient of each channel is positively correlated with the characteristic importance of the corresponding channel;
the second characteristic acquisition module is used for acquiring a target characteristic map of the vehicle image according to the multi-channel characteristic map and the weight coefficient of each channel;
and the classification module is used for acquiring the image category of the vehicle image based on the target feature map.
In a third aspect, an embodiment of the present application provides a computer device, which includes a memory and a processor, where the memory stores a computer program, and the processor implements the steps of the method according to the first aspect when executing the computer program.
In a fourth aspect, embodiments of the present application provide a computer-readable storage medium, on which a computer program is stored, which, when executed by a processor, implements the steps of the method according to the first aspect as described above.
The beneficial effects brought by the technical scheme provided by the embodiment of the application at least comprise:
obtaining a vehicle image to be classified; acquiring a multi-channel characteristic diagram corresponding to the vehicle image and a weight coefficient of each channel by adopting an image classification model; the weight coefficient of each channel is positively correlated with the characteristic importance of the corresponding channel; acquiring a target characteristic diagram of the vehicle image according to the multi-channel characteristic diagram and the weight coefficient of each channel; acquiring the image category of the vehicle image based on the target feature map; therefore, the computer equipment combines the multi-channel feature map of the vehicle image and the weight coefficient of each channel to obtain the target feature map of the vehicle image, wherein the weight coefficient of each channel is positively correlated with the feature importance of the corresponding channel, namely the higher the feature importance is, the larger the weight coefficient of the corresponding channel is, and further the proportion of the important features in the target feature map is improved; the problem that in the traditional technology, the output important feature information is less when the features of the image are extracted, and the image classification accuracy is further reduced is solved. According to the vehicle image classification method and device, the vehicle images are classified based on the target characteristic graph, and the image classification accuracy of the vehicle images can be improved.
Drawings
FIG. 1 is a flowchart illustrating an image processing method according to an embodiment;
FIG. 2 is a flowchart illustrating an image processing method according to an embodiment;
FIG. 3 is a partial structural diagram of an image classification model according to an embodiment;
FIG. 4 is a diagram illustrating an overall structure of an image classification model according to an embodiment;
FIG. 5 is a flowchart illustrating an image processing method according to an embodiment;
FIG. 6 is a diagram illustrating a training process of an image classification model according to an embodiment;
FIG. 7 is a diagram illustrating a training process of an image classification model according to an embodiment;
FIG. 8 is a diagram illustrating a training process of an image classification model according to an embodiment;
fig. 9 is a block diagram of an image processing apparatus according to an embodiment;
FIG. 10 is an internal block diagram of a computer device provided in one embodiment.
Detailed Description
In order to make the objects, technical solutions and advantages of the present application more apparent, the present application is described in further detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the present application and are not intended to limit the present application.
The image processing method, the image processing device, the computer equipment and the computer readable storage medium provided by the embodiment of the application aim to solve the technical problems that in the traditional technology, when the image is subjected to feature extraction, the output important feature information is less, and the image classification accuracy is influenced. The following describes in detail the technical solutions of the present application and how the technical solutions of the present application solve the above technical problems by embodiments and with reference to the drawings. The following specific embodiments may be combined with each other, and details of the same or similar concepts or processes may not be repeated in some embodiments.
In the image processing method provided in the embodiments of the present application, the execution subject may be an image processing apparatus, and the image processing apparatus may be implemented as part or all of a computer device by software, hardware, or a combination of software and hardware. In the following method embodiments, the execution subject is a computer device, which may be a server; it is understood that the image processing method provided by the following method embodiments may also be applied to a terminal, and may also be applied to a system including the terminal and a server, and is implemented through interaction between the terminal and the server.
Referring to fig. 1, a flowchart of an image processing method according to an embodiment of the present application is shown. The embodiment relates to a specific implementation process for classifying vehicle images by combining a multi-channel feature map of the vehicle images and weight coefficients of each channel. As shown in fig. 1, the image processing method of the present embodiment may include the steps of:
step S100, a vehicle image to be classified is acquired.
In this embodiment, the vehicle image to be classified may be an image that requires intelligent classification by a computer device in a vehicle detection process. As an implementation manner, when a vehicle is subjected to vehicle inspection at a vehicle inspection station, image acquisition equipment acquires an image of the vehicle to obtain an original image, and computer equipment locates a target area in the original image by using a target detection algorithm to obtain a position frame of the target area, wherein the target area can be a chassis area, a car light area, a carriage area and the like of the vehicle; and the computer device intercepts a target area corresponding to the detected position frame from the original image to obtain a vehicle image.
In other embodiments, the vehicle image to be classified may also be an image stored locally by the computer device and requiring automatic classification by the computer device, which is not specifically limited herein.
And step S200, acquiring a multi-channel characteristic diagram corresponding to the vehicle image and a weight coefficient of each channel by adopting an image classification model.
Wherein the weight coefficient of each channel is positively correlated with the characteristic importance of the corresponding channel.
In this embodiment, the image classification model may be a vgg (visual Geometry group) network model, an AlexNet model, or a ResNet residual network model, and the like, which is not limited herein.
The computer device obtains a multi-channel feature map of the vehicle image by using the image classification model, wherein the size of the multi-channel feature map can be c x h w, c is the number of channels of the multi-channel feature map, h is the height of the multi-channel feature map, and w is the width of the multi-channel feature map.
As an embodiment, the computer device may obtain the weight coefficient of each channel, i.e., each feature channel, through SE Net (Squeeze-and-importance networks). Specifically, the computer device embeds the SE module into the image classification model, models the correlation among the channels through the SE module, and outputs the weight coefficients corresponding to the channels of the multi-channel feature map, wherein the weight coefficients of the channels are all in the (0,1) interval, and the sum of the weight coefficients of the channels is 1. For example, the number of channels of the multi-channel feature map is 3, and the computer device obtains the weight coefficients corresponding to the three channels through the SE module, which are 0.2, 0.5, and 0.3, respectively. The weight coefficient of each channel is positively correlated with the feature importance of the corresponding channel, and continuing with the above example, in the three-channel feature map, the feature importance of the channel corresponding to the weight coefficient of 0.5 is the highest, so that the features with high importance are enhanced and the features with low importance are suppressed.
And step S300, acquiring a target characteristic diagram of the vehicle image according to the multi-channel characteristic diagram and the weight coefficient of each channel.
And the computer equipment acquires a target characteristic diagram of the vehicle image according to the multi-channel characteristic diagram and the weight coefficient of each channel. As an embodiment, the computer device may multiply each channel of the multi-channel feature map by the corresponding weight coefficient, that is, weight the weight coefficient of each channel onto the corresponding channel feature of the multi-channel feature map to obtain the target feature map.
And step S400, acquiring the image type of the vehicle image based on the target characteristic diagram.
And the computer equipment classifies the target characteristic graph by adopting an image classification model to obtain the image category corresponding to the vehicle image. For example, the vehicle image includes a chassis region, the corresponding image category may be a rectangle, square or circle, and the shapes of the different categories are used to represent different predicted shapes of the vehicle chassis dynamic identifier; the vehicle image comprises a vehicle lamp area, and the corresponding image category can be vehicle lamp on or vehicle lamp off; the vehicle image includes a carriage area, and the corresponding image category may be carriage normal or carriage abnormal, and the like.
The embodiment obtains the vehicle image to be classified; acquiring a multi-channel characteristic diagram corresponding to the vehicle image and a weight coefficient of each channel by adopting an image classification model; the weight coefficient of each channel is positively correlated with the characteristic importance of the corresponding channel; acquiring a target characteristic diagram of the vehicle image according to the multi-channel characteristic diagram and the weight coefficient of each channel; acquiring the image category of the vehicle image based on the target feature map; therefore, the computer device combines the multi-channel feature map of the vehicle image and the weight coefficient of each channel to obtain the target feature map of the vehicle image, wherein the weight coefficient of each channel is positively correlated with the feature importance of the corresponding channel, namely the higher the feature importance is, the larger the weight coefficient of the corresponding channel is, and the proportion of the important features in the target feature map is improved; the problem that in the traditional technology, the output important feature information is less when the features of the image are extracted, and the image classification accuracy is further reduced is solved. The vehicle image classification method and the vehicle image classification device classify the vehicle image based on the target characteristic diagram, and the image classification accuracy of the vehicle image can be improved.
In the practical application process, the computer equipment often needs to classify a plurality of vehicle images of different categories, and the resolution ratios of different vehicle images are different; because of the existence of the pooling layer in the image classification models such as the VGG network model, the AlexNet model or the ResNet residual error network model, the size of the output features is continuously reduced in the feature extraction process, and particularly for images with low resolution, the finally output useful feature information is less, so that the classification accuracy of the images is low; in this embodiment, the features with high importance are enhanced and the features with low importance are suppressed through the weight coefficients of the channels, so that the classification accuracy of the vehicle image, especially the vehicle image with low resolution, by the computer device can be enhanced.
Fig. 2 is a schematic flowchart of an image processing method according to another embodiment. On the basis of the embodiment shown in fig. 1, as shown in fig. 2, in the present embodiment, the step S300 includes a step S310 and a step S320, specifically:
and step S310, multiplying each channel of the multi-channel feature map by the corresponding weight coefficient respectively to obtain the multi-channel feature map after weight adjustment.
After the computer equipment acquires the vehicle images to be classified, the image classification model is adopted to acquire the multi-channel characteristic graphs corresponding to the vehicle images and the weight coefficients of all channels. In this embodiment, as an implementation manner, the image classification model is a residual error network model, specifically, a ResNet-18 residual error network model.
Specifically, referring to fig. 3, fig. 3 is a schematic view of a partial structure of the image classification model according to this embodiment. The partial structural diagram is a structural diagram of a residual block in the image classification model. The computer device embeds the SE module into the Residual network model, and as shown in fig. 3, the computer device performs feature compression on a multi-channel feature map c h w output by a Residual block Residual through a Global pooling layer (Global pooling) to obtain a multi-channel feature map c h 1, performs channel compression through a Fully connected layer (FC), and raises the channel back to the Global pooled dimension through a Fully connected layer, so that the model parameters and the calculated amount can be greatly reduced through two Fully connected layers, and finally obtains the weight coefficients corresponding to the channels of the multi-channel feature map through a sigmoid gate.
In this embodiment, the computer device multiplies each channel of the multi-channel feature map output by the Residual block Residual by the corresponding weight coefficient to obtain the weight-adjusted multi-channel feature map, specifically, weights the weight coefficient of each channel to the channel feature corresponding to the multi-channel feature map by a Scale function to obtain the weight-adjusted multi-channel feature map.
And step S320, correspondingly adding the pixel value of each pixel point of the vehicle image and the pixel value of each pixel point of the multichannel feature map after the weight adjustment to obtain a target feature map.
And correspondingly adding the pixel value of each pixel point of the vehicle image and the pixel value of each pixel point of the multichannel characteristic diagram after the weight adjustment by the computer equipment to obtain a target characteristic diagram. Referring to fig. 3, after obtaining the weighted multi-channel feature map by multiplying each channel of the multi-channel feature map by the corresponding weight coefficient through the Scale function, the computer device correspondingly adds the pixel value of each pixel point of the vehicle image x and the pixel value of each pixel point of the weighted multi-channel feature map to obtain the target feature map x —.
The computer device acquires an image category of the vehicle image based on the target feature map.
Specifically, referring to fig. 4, fig. 4 is a schematic diagram of an overall structure of the image classification model according to this embodiment. The computer equipment embeds an SE module in each residual block (Res2a, Res2b.. Res5b) of the image classification model, performs feature extraction on the vehicle image through each residual block embedded with the SE module to obtain a target feature map, and then obtains the image category of the vehicle image through the maximum pooling layer and the full connection layer of the image classification model.
Therefore, the SE module is embedded into each residual block of the image classification model, the features with high importance are strengthened through the weight coefficient of each channel, the features with low importance are restrained, and the accuracy of the classification result of the vehicle image is improved by improving the proportion of the features with high importance in the target feature map.
Fig. 5 is a flowchart illustrating an image processing method according to another embodiment. On the basis of the embodiment shown in fig. 1, as shown in fig. 5, the image processing method of the present embodiment further includes a step S500, specifically:
and S500, adjusting the size of the vehicle image by adopting a nearest neighbor interpolation method according to the required input size of the image classification model to obtain the adjusted vehicle image.
In this embodiment, after obtaining the vehicle image to be classified, the computer device inputs the size according to the requirement of the image classification model, and adjusts the size of the vehicle image by using a nearest neighbor interpolation method, so as to obtain an adjusted vehicle image. Specifically, the required input size of the image classification model is fixed, and is generally 224 × 224 or 64 × 64, and in the conventional technology, directly scaling an image with a larger length/width or a lower resolution to the required input size may cause image distortion and lose more effective information.
In this embodiment, in order to avoid the above problem, if the resolution of the vehicle image is lower than the required input size of the image classification model, the computer device adjusts the size of the vehicle image by using nearest neighbor interpolation. Specifically, the computer device sets the pixel value of the pixel point to be found as the nearest point in the vehicle image, for example, the vehicle image is an image of 4 × 4, the required input size of the image classification model is 8 × 8, the computer device first generates a blank image of 8 × 8, then fills the pixel value of the vehicle image at the image scaling position, fills the unfilled region of the image with the pixel value of the position nearest to the vehicle image, and obtains the adjusted vehicle image of 8 × 8 in size after filling.
Correspondingly, in this embodiment, step S200 includes step S210, specifically:
and step S210, acquiring a multi-channel feature map corresponding to the adjusted vehicle image and a weight coefficient of each channel by using an image classification model.
In this embodiment, the computer device adjusts the size of the vehicle image by using a nearest neighbor interpolation method according to the required input size of the image classification model to obtain an adjusted vehicle image, then obtains a multi-channel feature map corresponding to the adjusted vehicle image and a weight coefficient of each channel by using the image classification model, obtains a target feature map of the vehicle image according to the multi-channel feature map and the weight coefficient of each channel, and obtains the image category of the vehicle image based on the target feature map. Therefore, the problems of vehicle image distortion and feature loss caused by directly scaling the vehicle image with lower resolution to the required input size of the image classification model are solved, the embodiment ensures that the vehicle image is not distorted by adopting the nearest neighbor interpolation method, and the classification accuracy of the vehicle image is improved.
On the basis of the embodiment shown in fig. 1, referring to fig. 6, fig. 6 is a flowchart illustrating a training process of an image classification model according to another embodiment. As shown in fig. 6, the training process of the image classification model of this embodiment includes step S610 and step S620, specifically:
in step S610, a plurality of sample vehicle images are acquired.
In this embodiment, the computer device may acquire a plurality of sample vehicle images based on an actual application scenario, where the sample vehicle images may be images of a vehicle acquired by the computer device through the image acquisition device, or may be captured by the computer device from images of the vehicle acquired by the image acquisition device. The different sample vehicle images may include different areas of the vehicle, such as a chassis area, a headlight area, a bed area, and so forth.
The computer equipment classifies the sample vehicle images according to the corresponding categories respectively and stores the sample vehicle images in different folders, and the names of the folders are corresponding category labels.
As an embodiment, the computer device may label the plurality of category labels with different numbers, respectively. For example, the category label is marked with a number 0 to be a rectangle for the vehicle chassis dynamic identifier, a number 1 to be a square for the vehicle chassis dynamic identifier, a number 2 to be a vehicle light on, a number 3 to be a vehicle light off, and so on.
And S620, replacing the loss function of the initial residual error network model with a focal loss function, and training the replaced initial residual error network model according to a plurality of sample vehicle images to obtain an image classification model.
In this embodiment, the computer device obtains the image classification model by training using a residual error network model framework. The initial residual error network model is a parameter initialized residual error network model.
In the conventional technology, because a plurality of sample vehicle images comprise samples of different types, and the proportions of the samples of the different types may be unbalanced or even severely unbalanced, the image classification model cannot learn the features of the sample vehicle images of the types with smaller proportions in the training process.
In order to solve the problem that the image classification model cannot learn the characteristics of the sample vehicle images of the smaller-scale class due to the unbalanced sample classes, the computer device replaces the loss function softmax of the initial residual error network model with the focal loss function focal.
Referring to equation 1, equation 1 is the equation for the class two softmax loss:
where y represents the class of a sample vehicle image, and y' represents the output class of the sample vehicle image after softmax.
Referring to equation 2, equation 2 is the equation for Focal loss:
wherein, Lflα represents a balance factor for balancing the non-uniform proportion of the sample vehicle images of different classes, for example, α is 0.4 and r is a constant greater than 0 to reduce the loss of simple samples that are easy to classify, so that the training is more focused on difficult samples that are difficult to classify, for example, r is 2 in this embodiment.
Therefore, by comparing, the softmax loss of the initial residual error network model is replaced by the focal loss, so that the problem that the proportions of the vehicle images of different types of samples are seriously unbalanced in the training process can be avoided, the weight of a large amount of simple samples lost in the training is reduced, and the training is more concentrated on difficult samples which are difficult to correctly classify. It can be understood that focalloss can be applied to a two-classification task or a multi-classification task, thereby improving the classification accuracy of the image classification model.
Referring to fig. 7, fig. 7 is a flowchart illustrating a training process of an image classification model according to another embodiment. On the basis of the embodiment shown in fig. 6, as shown in fig. 7, in the training process of the image classification model of this embodiment, step S620 includes step S621, step S622, and step S623, specifically:
step S621, inputting the plurality of sample vehicle images into the replaced initial residual error network model, and obtaining an output category of each sample vehicle image.
And after obtaining the plurality of sample vehicle images, the computer equipment inputs the plurality of sample vehicle images into the replaced initial residual error network model to obtain the output category of each sample vehicle image.
In step S622, the label smoothing process is performed on the output type of each sample vehicle image according to the output type of each sample vehicle image.
In the conventional technology, when cross entropy (cross entropy) is used for the classification loss function, the true class prediction probability corresponding to the sample is fitted to the true probability, and the true class prediction score is much larger than other class scores, and Focalloss is just a modification performed on the basis of the cross entropy loss function. This may result in the image classification model tending to assign 1 to the true category prediction score of each sample vehicle image, easily resulting in overfitting; and the image classification model tends to make the difference in the prediction scores between the true class and the other classes as large as possible, which may cause the trained image classification model to believe too much the predicted class.
In this embodiment, in order to avoid the above problem, the computer device performs the label smoothing process on the output category of each sample vehicle image according to the output category of each sample vehicle image.
The formula of label smoothing is shown in formula 3:
new _ onehot _ labels (1-label _ smoothening) + label _ smoothening/num _ classes formula 3
The new _ onehot _ labels represents the output class after the label smoothing processing, the onehot _ labels represents the output class before the label smoothing processing of the corresponding sample vehicle image, namely the output class of the sample vehicle image output by the initial residual error network model, the label _ smoothing represents the smoothing coefficient, and the num _ classes represents the total class number of the classification.
For example, the total number of classes is 3, and in a case where the vehicle image belongs to class 2, the output class before the tag smoothing process is [0,1,0], the smoothing coefficient label _ smoothing is taken to be 0.1, and the output class after the tag smoothing process is [0,1,0] (1-0.1) +0.1/3 ═ 0,0.9,0] + [0.0333,0.0333,0.0333, 0.0333] - [0.0333,0.9333,0.0333 ].
Step S623, adjusting parameters of the initial residual error network model according to the output category after the label smoothing processing corresponding to each sample vehicle image.
And the computer equipment takes the output category after the label smoothing processing corresponding to each sample vehicle image as the classification category corresponding to each sample vehicle image, adjusts the parameters of the initial residual error network model, and performs iterative training to obtain an image classification model.
Referring to fig. 8, fig. 8 is a flowchart illustrating a training process of an image classification model according to another embodiment. On the basis of the embodiment shown in fig. 6, as shown in fig. 8, in the training process of the image classification model of this embodiment, step S610 includes step S611, step S612, and step S613, specifically:
in step S611, a plurality of original sample images are acquired.
The computer device acquires a plurality of original sample images, and the original sample images can be acquired by the computer device in an actual application scene. It can be understood that the original sample image may have the problems of unbalanced classes and less sample amount due to environmental limitations, which all reduce the model effect of the image classification model.
Step S612, performing data amplification processing on the plurality of original sample images to obtain a plurality of amplified sample images.
In this embodiment, in order to avoid the above problem, the computer device performs data amplification processing on the plurality of original sample images to obtain a plurality of amplified sample images. Specifically, the computer device performs multi-angle (-5, -10, 5,10) rotation, mirroring, Random cropping, Random Erasing operation, and the like on the original sample image to obtain a plurality of amplified sample images.
Step S613, adjusting the sizes of the plurality of amplified sample images by using a nearest neighbor interpolation method, to obtain a plurality of sample vehicle images.
For the amplified sample images with lower resolution, in order to avoid image distortion, the computer equipment adjusts the sizes of the multiple amplified sample images by adopting a nearest neighbor interpolation method, so that the sample vehicle images can meet the required input size of the residual error network model and can be ensured not to be distorted.
The computer device replaces the loss function of the initial residual error network model with a focal loss function, generates a training set and a verification set according to a plurality of sample vehicle images, trains the replaced initial residual error network model, and obtains an image classification model. Therefore, the problem of low classification accuracy caused by unbalanced training sample classes is solved, and the classification accuracy of the image classification model on the vehicle images is improved.
In practical applications, a computer device often needs to classify a plurality of vehicle images of different categories, and the different vehicle images have large differences. The image processing method provided by the above embodiment is based on an original image classification model, such as a ResNet-18 residual network model, and is respectively improved from three aspects of data preprocessing, a network model structure and a loss function, for the situations that the vehicle image categories are unbalanced and the vehicle image resolution and the aspect ratio are different greatly. Experimental results show that the image processing method in the embodiment can achieve ideal classification accuracy (> 95%) in each vehicle image category, and meets practical application requirements of projects.
It should be understood that, although the steps in the above-described flowcharts are shown in order as indicated by the arrows, the steps are not necessarily performed in order as indicated by the arrows. The steps are not performed in the exact order shown and described, and may be performed in other orders, unless explicitly stated otherwise. Moreover, at least a portion of the steps in the above-described flowcharts may include multiple sub-steps or multiple stages, which are not necessarily performed at the same time, but may be performed at different times, and the order of performing the sub-steps or the stages is not necessarily sequential, but may be performed alternately or alternatingly with other steps or at least a portion of the sub-steps or stages of other steps.
In one embodiment, as shown in fig. 9, there is provided an image processing apparatus including:
the image acquisition module 10 is used for acquiring a vehicle image to be classified;
the first feature obtaining module 20 is configured to obtain a multi-channel feature map corresponding to the vehicle image and a weight coefficient of each channel by using an image classification model; the weight coefficient of each channel is positively correlated with the characteristic importance of the corresponding channel;
a second feature obtaining module 30, configured to obtain a target feature map of the vehicle image according to the multi-channel feature map and the weight coefficients of the channels;
and the classification module 40 is used for acquiring the image category of the vehicle image based on the target feature map.
Optionally, the second feature obtaining module 30 includes:
the weight adjusting submodule is used for multiplying each channel of the multichannel feature map by a corresponding weight coefficient respectively to obtain a multichannel feature map after weight adjustment;
and the target characteristic map obtaining submodule is used for correspondingly adding the pixel value of each pixel point of the vehicle image and the pixel value of each pixel point of the multichannel characteristic map after the weight adjustment to obtain the target characteristic map.
Optionally, the image classification model is a residual network model.
Optionally, the apparatus further comprises:
the size adjusting module is used for adjusting the size of the vehicle image by adopting a nearest neighbor interpolation method according to the required input size of the image classification model to obtain an adjusted vehicle image;
correspondingly, the first feature obtaining module 20 includes:
and the characteristic obtaining submodule is used for obtaining a multi-channel characteristic diagram corresponding to the adjusted vehicle image and the weight coefficient of each channel by adopting the image classification model.
Optionally, the apparatus further comprises:
a sample acquisition module for acquiring a plurality of sample vehicle images;
and the training module is used for replacing the loss function of the initial residual error network model with a focus loss function focalloss, and training the replaced initial residual error network model according to the plurality of sample vehicle images to obtain the image classification model.
Optionally, the training module comprises:
the first training submodule is used for inputting the plurality of sample vehicle images into the replaced initial residual error network model to obtain the output category of each sample vehicle image;
the label smoothing sub-module is used for performing label smoothing on the output category of each sample vehicle image according to the output category of each sample vehicle image;
and the second training submodule is used for adjusting the parameters of the initial residual error network model according to the output category after the label smoothing processing corresponding to each sample vehicle image.
Optionally, the sample acquiring module includes:
the first sample acquisition submodule is used for acquiring a plurality of original sample images;
the sample amplification submodule is used for carrying out data amplification processing on the plurality of original sample images to obtain a plurality of amplified sample images;
and the second sample acquisition submodule is used for adjusting the sizes of the plurality of amplified sample images by adopting a nearest neighbor interpolation method to obtain the plurality of sample vehicle images.
The image processing apparatus provided in this embodiment may implement the above-mentioned embodiment of the image processing method, and the implementation principle and the technical effect are similar, which are not described herein again. For specific limitations of the image processing apparatus, reference may be made to the above limitations of the image processing method, which are not described herein again. The respective modules in the image processing apparatus described above may be wholly or partially implemented by software, hardware, and a combination thereof. The modules can be embedded in a hardware form or independent from a processor in the computer device, and can also be stored in a memory in the computer device in a software form, so that the processor can call and execute operations corresponding to the modules.
In one embodiment, there is also provided a computer device as shown in fig. 10, which may be a server. The computer device includes a processor, a memory, a network interface, and a database connected by a system bus. Wherein the processor of the computer device is configured to provide computing and control capabilities. The memory of the computer device comprises a nonvolatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program, and a database. The internal memory provides an environment for the operation of an operating system and computer programs in the non-volatile storage medium. The database of the computer device is used for storing image processing data. The network interface of the computer device is used for communicating with an external terminal through a network connection. The computer program is executed by a processor to implement an image processing method.
Those skilled in the art will appreciate that the architecture shown in fig. 10 is a block diagram of only a portion of the architecture associated with the subject application, and is not intended to limit the computing device to which the subject application may be applied, and that a computing device may in particular include more or less components than those shown, or combine certain components, or have a different arrangement of components.
In one embodiment, a computer device is provided, comprising a memory and a processor, the memory having a computer program stored therein, the processor implementing the following steps when executing the computer program:
acquiring a vehicle image to be classified;
acquiring a multi-channel characteristic diagram corresponding to the vehicle image and a weight coefficient of each channel by adopting an image classification model; the weight coefficient of each channel is positively correlated with the characteristic importance of the corresponding channel;
acquiring a target characteristic diagram of the vehicle image according to the multi-channel characteristic diagram and the weight coefficient of each channel;
and acquiring the image category of the vehicle image based on the target feature map.
In one embodiment, the processor, when executing the computer program, further performs the steps of:
multiplying each channel of the multichannel feature map by a corresponding weight coefficient respectively to obtain a weighted multichannel feature map;
and correspondingly adding the pixel value of each pixel point of the vehicle image and the pixel value of each pixel point of the multichannel characteristic diagram after the weight adjustment to obtain the target characteristic diagram.
In one embodiment, the image classification model is a residual network model.
In one embodiment, the processor, when executing the computer program, further performs the steps of:
according to the required input size of the image classification model, the size of the vehicle image is adjusted by adopting a nearest neighbor interpolation method to obtain an adjusted vehicle image;
correspondingly, the obtaining of the multi-channel feature map corresponding to the vehicle image and the weight coefficient of each channel by using the image classification model includes:
and acquiring a multi-channel characteristic diagram corresponding to the adjusted vehicle image and a weight coefficient of each channel by adopting the image classification model.
In one embodiment, the processor, when executing the computer program, further performs the steps of:
obtaining a plurality of sample vehicle images;
and replacing the loss function of the initial residual error network model with a focal loss function, and training the replaced initial residual error network model according to the plurality of sample vehicle images to obtain the image classification model.
In one embodiment, the processor, when executing the computer program, further performs the steps of:
inputting the plurality of sample vehicle images into the replaced initial residual error network model to obtain the output category of each sample vehicle image;
performing label smoothing processing on the output category of each sample vehicle image according to the output category of each sample vehicle image;
and adjusting the parameters of the initial residual error network model according to the output category after the label smoothing processing corresponding to each sample vehicle image.
In one embodiment, the window images include a front window image, a rear window image and a side window image, and the processor when executing the computer program further implements the following steps:
acquiring a plurality of original sample images;
performing data amplification processing on the plurality of original sample images to obtain a plurality of amplified sample images;
and adjusting the sizes of the plurality of amplified sample images by adopting a nearest neighbor interpolation method to obtain the plurality of sample vehicle images.
Those skilled in the art will appreciate that all or a portion of the processes in the methods of the embodiments described above may be implemented by hardware instructions associated with a computer program, which may be stored in a non-volatile computer-readable storage medium that, when executed, may include the processes of the embodiments of the methods described above, wherein any reference to memory, storage, database, or other medium used in the embodiments provided herein may include non-volatile and/or volatile memory, non-volatile memory may include read-only memory (ROM), programmable ROM (prom), electrically programmable ROM (eprom), electrically erasable programmable ROM (eeprom), or flash memory, volatile memory may include Random Access Memory (RAM) or external cache memory, and by way of illustration and not limitation, DRAM is available in a variety of forms, such as static RAM (sram), Dynamic RAM (DRAM), (sdram), synchronous DRAM, (sdram), dual data rate sdram), (dddram), (sdram), (rddram), and/DRAM).
In one embodiment, a computer-readable storage medium is provided, having a computer program stored thereon, which when executed by a processor, performs the steps of:
acquiring a vehicle image to be classified;
acquiring a multi-channel characteristic diagram corresponding to the vehicle image and a weight coefficient of each channel by adopting an image classification model; the weight coefficient of each channel is positively correlated with the characteristic importance of the corresponding channel;
acquiring a target characteristic diagram of the vehicle image according to the multi-channel characteristic diagram and the weight coefficient of each channel;
and acquiring the image category of the vehicle image based on the target feature map.
In one embodiment, the computer program when executed by the processor further performs the steps of:
multiplying each channel of the multichannel feature map by a corresponding weight coefficient respectively to obtain a weighted multichannel feature map;
and correspondingly adding the pixel value of each pixel point of the vehicle image and the pixel value of each pixel point of the multichannel characteristic diagram after the weight adjustment to obtain the target characteristic diagram.
In one embodiment, the image classification model is a residual network model.
In one embodiment, the computer program when executed by the processor further performs the steps of:
according to the required input size of the image classification model, the size of the vehicle image is adjusted by adopting a nearest neighbor interpolation method to obtain an adjusted vehicle image;
correspondingly, the obtaining of the multi-channel feature map corresponding to the vehicle image and the weight coefficient of each channel by using the image classification model includes:
and acquiring a multi-channel characteristic diagram corresponding to the adjusted vehicle image and a weight coefficient of each channel by adopting the image classification model.
In one embodiment, the computer program when executed by the processor further performs the steps of:
obtaining a plurality of sample vehicle images;
and replacing the loss function of the initial residual error network model with a focal loss function, and training the replaced initial residual error network model according to the plurality of sample vehicle images to obtain the image classification model.
In one embodiment, the computer program when executed by the processor further performs the steps of:
inputting the plurality of sample vehicle images into the replaced initial residual error network model to obtain the output category of each sample vehicle image;
performing label smoothing processing on the output category of each sample vehicle image according to the output category of each sample vehicle image;
and adjusting the parameters of the initial residual error network model according to the output category after the label smoothing processing corresponding to each sample vehicle image.
In one embodiment, the computer program when executed by the processor further performs the steps of:
acquiring a plurality of original sample images;
performing data amplification processing on the plurality of original sample images to obtain a plurality of amplified sample images;
and adjusting the sizes of the plurality of amplified sample images by adopting a nearest neighbor interpolation method to obtain the plurality of sample vehicle images.
The technical features of the above embodiments can be arbitrarily combined, and for the sake of brevity, all possible combinations of the technical features in the above embodiments are not described, but should be considered as the scope of the present specification as long as there is no contradiction between the combinations of the technical features.
The above examples only show some embodiments of the present invention, and the description thereof is more specific and detailed, but not construed as limiting the scope of the invention. It should be noted that, for a person skilled in the art, several variations and modifications can be made without departing from the inventive concept, which falls within the scope of the present invention. Therefore, the protection scope of the present patent shall be subject to the appended claims.
Claims (10)
1. An image processing method, characterized in that the method comprises:
acquiring a vehicle image to be classified;
acquiring a multi-channel characteristic diagram corresponding to the vehicle image and a weight coefficient of each channel by adopting an image classification model; the weight coefficient of each channel is positively correlated with the characteristic importance of the corresponding channel;
acquiring a target characteristic diagram of the vehicle image according to the multi-channel characteristic diagram and the weight coefficient of each channel;
and acquiring the image category of the vehicle image based on the target feature map.
2. The method according to claim 1, wherein the obtaining the target feature map of the vehicle image according to the multi-channel feature map and the weight coefficient of each channel comprises:
multiplying each channel of the multichannel feature map by a corresponding weight coefficient respectively to obtain a weighted multichannel feature map;
and correspondingly adding the pixel value of each pixel point of the vehicle image and the pixel value of each pixel point of the multichannel characteristic diagram after the weight adjustment to obtain the target characteristic diagram.
3. The method according to claim 1 or 2, wherein the image classification model is a residual network model.
4. The method of claim 1, wherein after the obtaining the image of the vehicle to be classified, further comprising:
according to the required input size of the image classification model, the size of the vehicle image is adjusted by adopting a nearest neighbor interpolation method to obtain an adjusted vehicle image;
correspondingly, the obtaining of the multi-channel feature map corresponding to the vehicle image and the weight coefficient of each channel by using the image classification model includes:
and acquiring a multi-channel characteristic diagram corresponding to the adjusted vehicle image and a weight coefficient of each channel by adopting the image classification model.
5. The method of claim 1, wherein the training process of the image classification model comprises:
obtaining a plurality of sample vehicle images;
and replacing the loss function of the initial residual error network model with a focal loss function, and training the replaced initial residual error network model according to the plurality of sample vehicle images to obtain the image classification model.
6. The method of claim 5, wherein training the replaced initial residual network model from the plurality of sample vehicle images to obtain the image classification model comprises:
inputting the plurality of sample vehicle images into the replaced initial residual error network model to obtain the output category of each sample vehicle image;
performing label smoothing processing on the output category of each sample vehicle image according to the output category of each sample vehicle image;
and adjusting the parameters of the initial residual error network model according to the output category after the label smoothing processing corresponding to each sample vehicle image.
7. The method of claim 5, wherein said obtaining a plurality of sample vehicle images comprises:
acquiring a plurality of original sample images;
performing data amplification processing on the plurality of original sample images to obtain a plurality of amplified sample images;
and adjusting the sizes of the plurality of amplified sample images by adopting a nearest neighbor interpolation method to obtain the plurality of sample vehicle images.
8. An image processing apparatus, characterized in that the apparatus comprises:
the image acquisition module is used for acquiring a vehicle image to be classified;
the first characteristic acquisition module is used for acquiring a multi-channel characteristic diagram corresponding to the vehicle image and a weight coefficient of each channel by adopting an image classification model; the weight coefficient of each channel is positively correlated with the characteristic importance of the corresponding channel;
the second characteristic acquisition module is used for acquiring a target characteristic map of the vehicle image according to the multi-channel characteristic map and the weight coefficient of each channel;
and the classification module is used for acquiring the image category of the vehicle image based on the target feature map.
9. A computer device comprising a memory and a processor, the memory storing a computer program, characterized in that the processor, when executing the computer program, implements the steps of the method of any of claims 1 to 7.
10. A computer-readable storage medium, on which a computer program is stored, which, when being executed by a processor, carries out the steps of the method of any one of claims 1 to 7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010306531.9A CN111488945A (en) | 2020-04-17 | 2020-04-17 | Image processing method, image processing device, computer equipment and computer readable storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010306531.9A CN111488945A (en) | 2020-04-17 | 2020-04-17 | Image processing method, image processing device, computer equipment and computer readable storage medium |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN111488945A true CN111488945A (en) | 2020-08-04 |
Family
ID=71795025
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010306531.9A Pending CN111488945A (en) | 2020-04-17 | 2020-04-17 | Image processing method, image processing device, computer equipment and computer readable storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111488945A (en) |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112287800A (en) * | 2020-10-23 | 2021-01-29 | 北京中科模识科技有限公司 | Advertisement video identification method and system under no-sample condition |
| CN112418283A (en) * | 2020-11-13 | 2021-02-26 | 三六零智慧科技(天津)有限公司 | Label smoothing method, device, equipment and storage medium for target detection |
| CN112784677A (en) * | 2020-12-04 | 2021-05-11 | 上海芯翌智能科技有限公司 | Model training method and device, storage medium and computing equipment |
| CN113255766A (en) * | 2021-05-25 | 2021-08-13 | 平安科技（深圳）有限公司 | Image classification method, device, equipment and storage medium |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109800795A (en) * | 2018-12-29 | 2019-05-24 | 广州市贺氏办公设备有限公司 | A kind of fruit and vegetable recognition method and system |
| CN109886964A (en) * | 2019-03-29 | 2019-06-14 | 北京百度网讯科技有限公司 | Circuit board defect detection method, device and equipment |
| CN110222732A (en) * | 2019-05-17 | 2019-09-10 | 上海工程技术大学 | A kind of vehicle checking method of multi-channel feature fusion |
| CN110494892A (en) * | 2017-05-31 | 2019-11-22 | 三星电子株式会社 | Method and apparatus for handling multi-channel feature figure image |
| CN110751162A (en) * | 2018-07-24 | 2020-02-04 | 杭州海康威视数字技术股份有限公司 | Image identification method and device and computer equipment |
| US20200074178A1 (en) * | 2018-08-31 | 2020-03-05 | Alibaba Group Holding Limited | Method and system for facilitating recognition of vehicle parts based on a neural network |
-
2020
- 2020-04-17 CN CN202010306531.9A patent/CN111488945A/en active Pending
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110494892A (en) * | 2017-05-31 | 2019-11-22 | 三星电子株式会社 | Method and apparatus for handling multi-channel feature figure image |
| CN110751162A (en) * | 2018-07-24 | 2020-02-04 | 杭州海康威视数字技术股份有限公司 | Image identification method and device and computer equipment |
| US20200074178A1 (en) * | 2018-08-31 | 2020-03-05 | Alibaba Group Holding Limited | Method and system for facilitating recognition of vehicle parts based on a neural network |
| CN109800795A (en) * | 2018-12-29 | 2019-05-24 | 广州市贺氏办公设备有限公司 | A kind of fruit and vegetable recognition method and system |
| CN109886964A (en) * | 2019-03-29 | 2019-06-14 | 北京百度网讯科技有限公司 | Circuit board defect detection method, device and equipment |
| CN110222732A (en) * | 2019-05-17 | 2019-09-10 | 上海工程技术大学 | A kind of vehicle checking method of multi-channel feature fusion |
Non-Patent Citations (3)
| Title |
|---|
| 周盛;王宵;王子豪;陶菁怡;邵叶秦;李泽慧;胡彬;陈锦花;王琼;: "基于多通道背景提取算法的车辆检测", no. 11, pages 209 - 212 * |
| 李德毅 等: "《中国科协新一代信息技术系列丛书 人工智能导论》", 31 August 2018, 中国科学技术出版社, pages: 160 - 161 * |
| 赵小川: "《MATLAB图像处理 程序实现与模块化仿真》", 31 January 2014, 北京航空航天大学出版社, pages: 78 - 79 * |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112287800A (en) * | 2020-10-23 | 2021-01-29 | 北京中科模识科技有限公司 | Advertisement video identification method and system under no-sample condition |
| CN112418283A (en) * | 2020-11-13 | 2021-02-26 | 三六零智慧科技(天津)有限公司 | Label smoothing method, device, equipment and storage medium for target detection |
| CN112418283B (en) * | 2020-11-13 | 2023-07-11 | 三六零智慧科技(天津)有限公司 | Label smoothing method, device, equipment and storage medium for target detection |
| CN112784677A (en) * | 2020-12-04 | 2021-05-11 | 上海芯翌智能科技有限公司 | Model training method and device, storage medium and computing equipment |
| CN113255766A (en) * | 2021-05-25 | 2021-08-13 | 平安科技（深圳）有限公司 | Image classification method, device, equipment and storage medium |
| CN113255766B (en) * | 2021-05-25 | 2023-12-22 | 平安科技（深圳）有限公司 | Image classification method, device, equipment and storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108961235B (en) | Defective insulator identification method based on YOLOv3 network and particle filter algorithm | |
| CN111488945A (en) | Image processing method, image processing device, computer equipment and computer readable storage medium | |
| CN110414507B (en) | License plate recognition method and device, computer equipment and storage medium | |
| CN113221905B (en) | Semantic segmentation unsupervised domain adaptation method, device and system based on uniform clustering and storage medium | |
| CN109753913B (en) | Multi-mode video semantic segmentation method with high calculation efficiency | |
| WO2023138300A1 (en) | Target detection method, and moving-target tracking method using same | |
| CN111797983A (en) | Neural network construction method and device | |
| CN109214403B (en) | Image recognition method, device and equipment and readable medium | |
| CN112116599B (en) | Sputum smear tubercle bacillus semantic segmentation method and system based on weak supervised learning | |
| CN113468967A (en) | Lane line detection method, device, equipment and medium based on attention mechanism | |
| CN111696110B (en) | Scene segmentation method and system | |
| CN111768415A (en) | Image instance segmentation method without quantization pooling | |
| CN112539843A (en) | Method and device for detecting temperature and computer equipment | |
| CN114926722A (en) | Method and storage medium for detecting scale self-adaptive target based on YOLOv5 | |
| CN111179270A (en) | Image co-segmentation method and device based on attention mechanism | |
| CN110008949B (en) | Image target detection method, system, device and storage medium | |
| CN109492697B (en) | Picture detection network training method and picture detection network training device | |
| CN111444923A (en) | Image semantic segmentation method and device under natural scene | |
| CN112926595B (en) | Training device of deep learning neural network model, target detection system and method | |
| CN113344791A (en) | Binocular super-resolution image detection method, system and medium based on cavity convolution and feature fusion | |
| CN112766351A (en) | Image quality evaluation method, system, computer equipment and storage medium | |
| CN115115552B (en) | Image correction model training method, image correction device and computer equipment | |
| CN112699809B (en) | Vaccinia category identification method, device, computer equipment and storage medium | |
| CN114549932A (en) | Data enhancement processing method and device, computer equipment and storage medium | |
| CN112949731A (en) | Target detection method, device, storage medium and equipment based on multi-expert model |
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 |