FuCiTNet: Improving the generalization of deep learning networks by the fusion of learned class-inherent transformations

TL;DR

FuCiTNet is a novel deep learning approach that enhances generalization on small datasets by learning class-specific transformations through multiple generators, inspired by GANs, leading to improved classification performance.

Contribution

The paper introduces FuCiTNet, a new method that learns class-inherent transformations to improve DNN generalization on small, visually similar datasets, independent of traditional techniques.

Findings

Transformations improve classification accuracy on small datasets

Method effective across diverse datasets

Generates class-specific transformations for better generalization

Abstract

It is widely known that very small datasets produce overfitting in Deep Neural Networks (DNNs), i.e., the network becomes highly biased to the data it has been trained on. This issue is often alleviated using transfer learning, regularization techniques and/or data augmentation. This work presents a new approach, independent but complementary to the previous mentioned techniques, for improving the generalization of DNNs on very small datasets in which the involved classes share many visual features. The proposed methodology, called FuCiTNet (Fusion Class inherent Transformations Network), inspired by GANs, creates as many generators as classes in the problem. Each generator, $k$, learns the transformations that bring the input image into the k-class domain. We introduce a classification loss in the generators to drive the leaning of specific k-class transformations. Our experiments demonstrate that the proposed transformations improve the generalization of the classification model in three diverse datasets.