An image feature extraction device based on small sample learning

Abstract

The present invention provides an image feature extraction device based on small sample learning, including a first multi-channel feature weighting module, a second multi-channel feature weighting module, a convolution module, a first multi-scale feature fusion module and a second multi-scale feature fusion module; the first multi-channel feature weighting module, the second multi-channel feature weighting module and the convolution module are sequentially connected; the first multi-scale feature fusion module is connected to the second multi-scale feature fusion module ; The first multi-scale feature fusion module is connected to the second multi-channel feature weighting module and the convolution module respectively; the second multi-scale feature fusion module is connected to the first multi-channel feature weighting module. The present invention extracts features from a small-sample image data set through a multi-channel feature weighting module and a multi-scale feature fusion module, which can effectively improve the accuracy of small-sample classification.

Description

technical field [0001] The invention relates to the technical field of image processing, in particular to an image feature extraction device based on small sample learning. Background technique [0002] In recent years, big data technology, convolutional neural network, and computing and performance have all developed rapidly. Large-scale data image tasks, such as image classification, object detection, and image segmentation, have all developed very maturely. However, the current mainstream deep learning network models are all proposed for tasks with a large number of samples, while ignoring the problem of scarcity of data in most task scenarios in real life. Compared with machines, humans can achieve fairly accurate learning results with only a small number of samples, mainly because humans have acquired the ability to quickly capture the distinguishing characteristics of samples during the long evolutionary process. Similarly, in the case of very scarce image samples, ho...

AI Technical Summary

Problems solved by technology

However, the current mainstream deep learning network models are all proposed for tasks with a large number of samples, while ignoring the problem of scarcity of data in most task scenarios in real life.

Images