3rd Place Solution to ICCV LargeFineFoodAI Retrieval

TL;DR

This paper presents a competitive solution for food image retrieval combining multiple models, advanced loss functions, test-time augmentation, ensemble techniques, and a novel reranking method, achieving top-tier performance in the ICCV LargeFineFoodAI challenge.

Contribution

The paper introduces a new reranking method based on diffusion and k-reciprocal reranking, enhancing retrieval accuracy in large-scale food image datasets.

Findings

Achieved 0.81219 mAP@100 on public leaderboard

Utilized weighted sum of ArcFace and Circle loss for training

Implemented diffusion-based reranking for improved retrieval

Abstract

This paper introduces the 3rd place solution to the ICCV LargeFineFoodAI Retrieval Competition on Kaggle. Four basic models are independently trained with the weighted sum of ArcFace and Circle loss, then TTA and Ensemble are successively applied to improve feature representation ability. In addition, a new reranking method for retrieval is proposed based on diffusion and k-reciprocal reranking. Finally, our method scored 0.81219 and 0.81191 mAP@100 on the public and private leaderboard, respectively.