Deep Video Generation, Prediction and Completion of Human Action Sequences Authors: Haotian Xue, Wenqing Chu, Zhao-Jie Ju, et al.

Mapping where a subject is looking while performing a task. Where to Access ArXiv: You can read the full methodology and results here.

The "g60104" identifier is part of the or GTEA Gaze+ datasets, which are used to evaluate computer vision models on:

Identifying specific tasks like "folding a tortilla" or "pouring water."

This work is frequently cited in the context of the KTH Action Dataset , where video files follow similar alphanumeric naming conventions. Context of the Video

Analyzing video filmed from a first-person (head-mounted) perspective.

If you are looking for the raw video file itself for testing, it is typically hosted on the Georgia Tech FovLab website under the GTEA dataset sections.