University of Bristol
Visual Information Laboratory

Bristol Egocentric Object Interaction Dataset (BEOID)

Further details of the entire dataset can be found at: http://www.cs.bris.ac.uk/~damen/BEOID/

File Format
==========
Each video within the dataset has a corresponding annotation file for annotated ACTIONS.

The data inside the annotation files is in comma separated value (csv) format and is arranged as follows:

Start Frame, End Frame,Action,Object(s)

Where:
--Start Frame is the first frame of the action.
--End Frame is the last frame of the action.
--Action is the annotated action chosen from WordNet represented as a synset. Further information about WordNet can be found at: https://wordnet.princeton.edu/
--Object(s) are any objects the user interacts with during the action. There can be any number of objects each separated by a comma.
