This challenge focuses on fine-grained action recognition from short first-person video clips.

Each clip captures a person interacting with objects in a kitchen environment. For every clip, participants must predict:
Verb
the action being performed
Noun
the object involved
This is a supervised multi-class classification problem with two prediction targets per clip. The objective is to build models that learn meaningful spatio-temporal representations and generalize to unseen environments.