Ablation study of self-supervised learning for image classification

TL;DR

This paper investigates the effectiveness of self-supervised learning methods for image classification using CNNs and transformers, evaluating their performance on multiple datasets.

Contribution

It presents an ablation study on self-supervised training techniques for CNNs and transformers, highlighting their impact on image recognition accuracy.

Findings

Self-supervised learning improves image classification performance.

Different backbones influence the effectiveness of the method.

Evaluation on three datasets demonstrates the method's robustness.

Abstract

This project focuses on the self-supervised training of convolutional neural networks (CNNs) and transformer networks for the task of image recognition. A simple siamese network with different backbones is used in order to maximize the similarity of two augmented transformed images from the same source image. In this way, the backbone is able to learn visual information without supervision. Finally, the method is evaluated on three image recognition datasets.