Exploiting Semantic Role Contextualized Video Features for Multi-Instance Text-Video Retrieval EPIC-KITCHENS-100 Multi-Instance Retrieval Challenge 2022
This is an incremental improvement for video retrieval tasks, specifically in kitchen activity datasets.
The paper tackled the problem of multi-instance text-video retrieval by parsing sentences into semantic roles and using self-attentions to contextualize video features, achieving a 3rd place ranking in nDCG and 4th in mAP on the EPIC-KITCHENS-100 challenge.
In this report, we present our approach for EPIC-KITCHENS-100 Multi-Instance Retrieval Challenge 2022. We first parse sentences into semantic roles corresponding to verbs and nouns; then utilize self-attentions to exploit semantic role contextualized video features along with textual features via triplet losses in multiple embedding spaces. Our method overpasses the strong baseline in normalized Discounted Cumulative Gain (nDCG), which is more valuable for semantic similarity. Our submission is ranked 3rd for nDCG and ranked 4th for mAP.