activity
20182025
most citedMTLE: A Multitask Learning Encoder of Visual Feature Representations for Video and Movie Description

3 citations · 3 across the 4 of their papers we have counts for

collaborators

4 papers