A Few-Sample Image Classification Method Based on Multi-decision Fusion

Abstract

The invention discloses a small-sample image classification method with multi-decision fusion, which belongs to the technical field of pattern recognition, and improves the validity and robustness of the model by comprehensively considering the decisions of multiple classifiers. This method is a simple non-parametric method, which can effectively solve the problem that the training data cannot be well adapted to the new category when the training data is limited. Classification performance for few-shot image classification.

Description

technical field [0001] The invention relates to the technical field of pattern recognition, in particular to a small-sample image classification method for multi-decision fusion. Background technique [0002] In recent years, deep learning, as a powerful tool, has helped machines reach or even surpass human performance in various visual tasks. One of the indispensable factors is large-scale labeled data. However, due to the limitations of actual situations, in the real world It may not be feasible to collect a large amount of labeled data in , so few-shot learning to solve this problem when labeled samples are scarce has attracted more and more attention. Currently popular small sample learning models usually include two parts: pre-training part: use the basic data to generate a neural network-based feature extraction model; testing phase part: first extract the embedded features of the test data, and then design a classifier to identify the query sample. [0003] At prese...

AI Technical Summary

Problems solved by technology

[0007] In order to solve the problems existing in the image classification process of the small sample image classification method in the prior art, the embodiment of the present invention provides a small sample image classification method with multi-decision fusion, which improves the model by comprehensively considering the decisions of multiple classifiers. The effectiveness and robustness, effectively solve the problem of not being able to adapt to new categories when the training data is limited, and greatly improve the classification performance of small-sample image classification

Images