Competing Ratio Loss for Discriminative Multi-class Image Classification

TL;DR

This paper introduces the Competing Ratio Loss (CRL), a new loss function for multi-class image classification that enhances class discrimination by considering the ratio of correct to wrong class probabilities, improving performance across various datasets.

Contribution

The paper proposes the novel Competing Ratio Loss (CRL) that directly compares correct and wrong class probabilities, addressing limitations of cross-entropy loss in discriminating classes.

Findings

CRL improves classification accuracy on multiple datasets.

CRL demonstrates robustness across different neural network architectures.

CRL enhances class separation in fine-grained and large-scale tasks.

Abstract

The development of deep convolutional neural network architecture is critical to the improvement of image classification task performance. A lot of studies of image classification based on deep convolutional neural network focus on the network structure to improve the image classification performance. Contrary to these studies, we focus on the loss function. Cross-entropy Loss (CEL) is widely used for training a multi-class classification deep convolutional neural network. While CEL has been successfully implemented in image classification tasks, it only focuses on the posterior probability of correct class when the labels of training images are one-hot. It cannot be discriminated against the classes not belong to correct class (wrong classes) directly. In order to solve the problem of CEL, we propose Competing Ratio Loss (CRL), which calculates the posterior probability ratio between the correct class and competing wrong classes to better discriminate the correct class from competing wrong classes, increasing the difference between the negative log likelihood of the correct class and the negative log likelihood of competing wrong classes, widening the difference between the probability of the correct class and the probabilities of wrong classes. To demonstrate the effectiveness of our loss function, we perform some sets of experiments on different types of image classification datasets, including CIFAR, SVHN, CUB200- 2011, Adience and ImageNet datasets. The experimental results show the effectiveness and robustness of our loss function on different deep convolutional neural network architectures and different image classification tasks, such as fine-grained image classification, hard face age estimation and large-scale image classification.