A kind of image recognition method and device including surrounding environment
A surrounding environment and image recognition technology, applied in the information field, can solve problems that do not involve image recognition
- Summary
- Abstract
- Description
- Claims
- Application Information
AI Technical Summary
Problems solved by technology
Method used
Image
Examples
Embodiment 1
[0059] In the prior art, the core convolution layer used in the residual network is pre-added with padding to the image. The input and output matrices of this convolution layer are of the same size, and the residual operation is performed F(x)=H(x)- x. Among them, F(X) is the residual function, which refers to the difference between the output of the module and the output, H(X) refers to the output matrix of the module, and X refers to the input matrix (image matrix) of the module.
[0060] In the prior art, only Chinese patent CN102694961B mentions the technical solution of "edge reduction", but the "edge reduction" in this patent refers to directly removing the pixels at the edge of the image after image processing. This is completely different from the definition of the narrow edge of the present invention, and the purpose of use is also irrelevant. The shrinking edge of the present invention refers to removing the characteristic elements of the matrix edge in the neuron n...
Embodiment 2
[0093] This embodiment is a further supplement and description to the foregoing embodiment, and repeated content will not be repeated.
[0094] This embodiment provides an image recognition device including a surrounding environment, characterized in that the device at least includes:
[0095] Convolution module for extracting image matrix based on convolution layer;
[0096] A shrinking unit, which is used to perform at least one shrinking process on the image matrix;
[0097] The residual unit is used to perform at least one non-reduced residual processing on the image matrix after the edge reduction processing;
[0098] The fully connected layer module is used to output the recognized image data based on the fully connected layer. Wherein, the edge reduction unit is arranged between the convolution module and the residual unit. That is, the convolution module is arranged at the data input end and connected to the edge reduction unit, so as to input the image matrix extra...
PUM
Login to View More Abstract
Description
Claims
Application Information
Login to View More 


