3rd Place: A Global and Local Dual Retrieval Solution to Facebook AI Image Similarity Challenge

TL;DR

This paper presents a dual global and local feature retrieval approach for image similarity, achieving third place in Facebook AI's challenge by combining optimized descriptors and robust matching techniques.

Contribution

It introduces a novel multi-branch retrieval method that integrates global and local descriptors with advanced optimization strategies for improved accuracy.

Findings

Global and local features are complementary in image retrieval.

Data augmentation and self-supervised learning enhance global descriptor performance.

SIFT features and GPU Faiss improve local retrieval robustness.

Abstract

As a basic task of computer vision, image similarity retrieval is facing the challenge of large-scale data and image copy attacks. This paper presents our 3rd place solution to the matching track of Image Similarity Challenge (ISC) 2021 organized by Facebook AI. We propose a multi-branch retrieval method of combining global descriptors and local descriptors to cover all attack cases. Specifically, we attempt many strategies to optimize global descriptors, including abundant data augmentations, self-supervised learning with a single Transformer model, overlay detection preprocessing. Moreover, we introduce the robust SIFT feature and GPU Faiss for local retrieval which makes up for the shortcomings of the global retrieval. Finally, KNN-matching algorithm is used to judge the match and merge scores. We show some ablation experiments of our method, which reveals the complementary advantages of global and local features.