0

Zero-shot Natural Language Video Localization

A zero-shot natural language video localization model is trained using pseudo-supervision generated from unpaired video and text data, outperforming supervised methods on Charades-STA and ActivityNet-Captions.

Year
2021
Venue
ICCV 2021 10
Authors
5
Hosting
Abstract onlyARXIV-DEFAULT

Cite

Notes

Only stored in your browser.

Attribution

Abstract & full text
arxiv.org/abs/2110.00428ARXIV-DEFAULT
TL;DR
Semantic Scholar
Attribution policy →

Abstract

Understanding videos to localize moments with natural language often requires large expensive annotated video regions paired with language queries. To eliminate the annotation costs, we make a first attempt to train a natural language video localization model in zero-shot manner. Inspired by unsupervised image captioning setup, we merely require random text corpora, unlabeled video collections, and an off-the-shelf object detector to train a model. With the unpaired data, we propose to generate pseudo-supervision of candidate temporal regions and corresponding query sentences, and develop a simple NLVL model to train with the pseudo-supervision. Our empirical validations show that the proposed pseudo-supervised method outperforms several baseline approaches and a number of methods using stronger supervision on Charades-STA and ActivityNet-Captions.

Authors

5