Results and findings of the 2021 Image Similarity Challenge

TL;DR

This paper analyzes the 2021 Image Similarity Challenge, highlighting the dataset, top methods, and challenges like severe crops and overlays, with insights into effective algorithmic strategies.

Contribution

It provides a comprehensive analysis of top submissions and identifies key techniques that improve image copy detection performance.

Findings

Severe crops and overlays are the most challenging transformations.

Training with strong augmentations and self-supervised learning improves results.

Explicit overlay detection and global descriptor matching are effective strategies.

Abstract

The 2021 Image Similarity Challenge introduced a dataset to serve as a new benchmark to evaluate recent image copy detection methods. There were 200 participants to the competition. This paper presents a quantitative and qualitative analysis of the top submissions. It appears that the most difficult image transformations involve either severe image crops or hiding into unrelated images, combined with local pixel perturbations. The key algorithmic elements in the winning submissions are: training on strong augmentations, self-supervised learning, score normalization, explicit overlay detection, and global descriptor matching followed by pairwise image comparison.