Image similarity using Deep CNN and Curriculum Learning

TL;DR

This paper introduces SimNet, a deep siamese network trained with curriculum learning for image similarity, utilizing multi-scale CNN embeddings to improve fine-grained similarity detection.

Contribution

The paper presents a novel online pair mining strategy and a multi-scale CNN architecture that enhances image similarity measurement over traditional methods.

Findings

Multi-scale siamese network outperforms traditional CNNs in capturing fine-grained similarities.

Curriculum learning improves the training efficiency and effectiveness of the siamese network.

Online pair mining strategy effectively selects training pairs for better model performance.

Abstract

Image similarity involves fetching similar looking images given a reference image. Our solution called SimNet, is a deep siamese network which is trained on pairs of positive and negative images using a novel online pair mining strategy inspired by Curriculum learning. We also created a multi-scale CNN, where the final image embedding is a joint representation of top as well as lower layer embedding's. We go on to show that this multi-scale siamese network is better at capturing fine grained image similarities than traditional CNN's.