/AI8h ago

CVPR Research Bridges Gap Between Captions And User Queries In Video Retrieval

1300779
#1728Dima Damen @CVPR
Original post
Dima Damen @CVPR@dimadamen#1728inAI

We identify a significant domain gap between the descriptive captions used in VMR training and the actual search patterns of real users. We design a pipeline to adjust queries in Video Moment Retrieval (VMR) benchmarks to emulate the linguistic distribution of user queries 2/3

Dima Damen @CVPR@dimadamen

Today @CVPR Poster Session 3 from 11:45, come to #400 for our poster Beyond Caption-Based Queries for Video Moment Retrieval work led by @pujolpe14 w/ @mwray0 @SergioEscalera_ &A Calpes @bristolcs @UniBarcelona 馃У 1/3

9:50 AM 路 Jun 6, 2026 路 518 Views
Sentiment
Sentiment building, check back later.
Cluster Engagement
Posts from X
Most Activity
Most Activity
VIEWS261LIKES1
Dima Damen @CVPR@dimadamen

We quantify and recover the performance of VMR models as query specificity changes over three datasets: HD-EPIC, YouCook2 and Activity-Net Captions.

Come talk to us at poster #400 @CVPR

Dima Damen @CVPR@dimadamen

We identify a significant domain gap between the descriptive captions used in VMR training and the actual search patterns of real users. We design a pipeline to adjust queries in Video Moment Retrieval (VMR) benchmarks to emulate the linguistic distribution of user queries 2/3

8hViews 261Likes 1Bookmarks 0