Dataset Groups Activity Stream CrossTask The CrossTask dataset contains 2,750 instructional videos, annotated for 133 keystep labels spanning 18 tasks. BibTex: @dataset{Dimitri_Zhukov_and_Jean-Baptiste_Alayrac_and_Ramazan_Gokberk_Cinbis_and_David_Fouhey_and_Ivan_Laptev_and_Josef_Sivic_2024, abstract = {The CrossTask dataset contains 2,750 instructional videos, annotated for 133 keystep labels spanning 18 tasks.}, author = {Dimitri Zhukov and Jean-Baptiste Alayrac and Ramazan Gokberk Cinbis and David Fouhey and Ivan Laptev and Josef Sivic}, doi = {10.57702/jjm3k6sl}, institution = {No Organization}, keyword = {'CrossTask dataset', 'instructional videos', 'keystep recognition', 'step labels', 'video action localization'}, month = {dec}, publisher = {TIB}, title = {CrossTask}, url = {https://service.tib.eu/ldmservice/dataset/crosstask}, year = {2024} }
