Generative Adversarial Network-Based Zero-Shot Image Classification System and Its Method

Abstract

The invention relates to a zero-sample image classification model based on a generative confrontation network, comprising a generative confrontation network module for obtaining visual error information; a visual feature extraction network processing module for obtaining a one-dimensional visual feature vector of an image; an attribute semantic conversion network module, which uses two-layer linear activation layers to map low-dimensional attribute semantic vectors to high-dimensional feature vectors with the same dimension as visual feature vectors; the visual-attribute semantic connection network realizes the fusion of visual feature vectors and attribute semantic feature vectors; score The classification result and reward output module uses cross-entropy loss to classify the seen labeled categories, and the reward output is to penalize the unseen unlabeled data, as well as penalize the seen and unseen categories in the unlabeled data The most likely prediction result for the class that has been passed. The invention can effectively solve the problem of missing image category labels.

Description

technical field [0001] The invention relates to a zero-sample image classification model, in particular to a zero-sample image classification system based on a generative confrontation network and a method thereof. Background technique [0002] Currently, in the process of image classification, if you want to accurately classify images, you need to inform the model of the image labels of each category. However, the number of image categories is often very large, and new categories may be added from time to time. If each category label is manually labeled each time, the workload will be extremely huge. In this process, some categories have only a few or no training sample labels, and the entire categories without training labels belong to zero samples. Such zero samples cannot be used to construct classifiers using traditional machine learning methods. The purpose of zero-shot learning image classification is to solve the problem of missing the entire category label and clas...

AI Technical Summary

Problems solved by technology

However, the number of image categories is often very large, and new categories may be added from time to time. If each category label is manually labeled each time, the workload will be extremely huge.

In this process, some categories have only a few or no training sample labels, and the entire category without training labels belongs to zero samples. Such zero samples cannot be effectively constructed by using traditional machine learning methods.

Images