Fine-grained image classification method based on feature pyramid and global average pooling

Abstract

The invention discloses a fine-grained image classification method based on feature pyramid and global average pooling, comprising the following steps: step 1, image input into the convolutional layer of a pre-trained convolutional neural network to obtain a multi-channel feature map; step 2 , the feature map of the multi-channel passes through the global average pooling layer to obtain the saliency map of the input image, and extract the position information of the target; step 3, the feature pyramid network extracts the features of the feature map of the multi-channel and predicts to obtain the largest amount of information K local areas; step 4, aggregating the local features of the K local areas and the global features of the input image through the convolutional neural network to predict and output the final recognition category. The method of the invention reduces the influence of background noise, enhances the robustness of local area selection, and improves the recognition accuracy.

Description

technical field [0001] The invention relates to a fine-grained image classification method, in particular to a fine-grained image classification method based on feature pyramid and global average pooling. Background technique [0002] Fine-grained image recognition is a concept in the field of image processing. Traditional image recognition can only recognize the large categories of objects in the image, which is called coarse-grained image recognition. However, there are usually many subcategories under the same category, and traditional image recognition methods cannot determine the specific subcategory to which the target belongs. Fine-grained image recognition can classify the targets in the image more finely, and the granularity of the classification is finer. It is required to determine the specific subcategory of the target under the large category to meet the requirements of higher image recognition in different scenarios. Require. [0003] Early fine-grained class...

AI Technical Summary

Problems solved by technology

[0008] In view of the above-mentioned defects in the prior art, the task of the present invention is to provide a fine-grained image classification method based on feature pyramid and global average pooling, which uses less computational overhead to solve the noise problem in the target positioning area, and improves the efficiency of target object extraction. feature robustness

Images