UniUD-FBK-UB-UniBZ Submission to the EPIC-Kitchens-100 Multi-Instance Retrieval Challenge 2022

TL;DR

This paper details a model ensemble approach using relevance-augmented triplet loss for the EPIC-Kitchens-100 retrieval challenge, achieving competitive nDCG and mAP scores.

Contribution

The paper introduces a novel ensemble of models trained with relevance-augmented triplet loss variants for multi-instance retrieval.

Findings

Achieved 61.02% nDCG on the leaderboard

Achieved 49.77% mAP on the leaderboard

Demonstrated effectiveness of relevance-augmented triplet loss

Abstract

This report presents the technical details of our submission to the EPIC-Kitchens-100 Multi-Instance Retrieval Challenge 2022. To participate in the challenge, we designed an ensemble consisting of different models trained with two recently developed relevance-augmented versions of the widely used triplet loss. Our submission, visible on the public leaderboard, obtains an average score of 61.02% nDCG and 49.77% mAP.