Our new X account is live! Follow @wizwand_team for updates
WorkDL logo mark

Zero-shot Natural Language Video Localization

About

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.

Jinwoo Nam, Daechul Ahn, Dongyeop Kang, Seong Jong Ha, Jonghyun Choi• 2021

Related benchmarks

TaskDatasetResultRank
Temporal Video GroundingCharades-STA (test)
Recall@IoU=0.531.29
117
Video Moment RetrievalCharades-STA
R1@0.530.9
44
Temporal Video GroundingActivityNet-Captions (test)
Recall@IoU>0.344.74
22
Showing 3 of 3 rows

Other info

Follow for update