Fine-grained image classification method based on block convolutional neural network

TL;DR: In this article, a fine-grained image classification method based on a block convolutional neural network was proposed, where extra parameters and operations are not introduced, so the high efficiency of a general convolution neural network is reserved in the prediction process, the input feature map does not need to be blocked by the characteristics of an overlarge receptive field, each block is subjected to convolution operation and then spliced again, and the method has high restrictionivity.

Abstract: The invention discloses a fine-grained image classification method based on a block convolutional neural network, relates to the technical field of fine-grained image recognition, and solves a problemthat the reception field limitation is weaker because in an existing method, an original image is averagely partitioned and then is input into a convolutional neural network for fine-grained image classification. According to the invention, extra parameters and operations are not introduced, so the high efficiency of a general convolutional neural network is reserved in the prediction process, the input feature map does not need to be blocked by the characteristics of an overlarge receptive field, each block is subjected to convolution operation and then spliced again, and the method has highrestrictivity. According to the method, the convolutional receptive field is limited as required, so that the network pays more attention to the characteristics of the local area, and the method is more suitable for fine-grained image classification tasks. According to the fine-grained image classification method, on the premise of not introducing more parameters, the receptive field range of theconvolutional layer is limited, so the convolutional neural network can find a small local area with discrimination ability.