romerobarata / hierarchical_action_prediction Goto Github PK

Hi Romero,

Thanks for sharing the code. In the paper you mentioned 2-stage training, does it mean we should run train.py two times, and in the second training loading the pre-trained model while changing parameters of the segment loss and pattern?

Thanks and looking forward to your reply.

Best,
Yutasq

Hi, thank you for the great project.
Does the input of HERA contain only text information (label and duration of action) without visual information (visual features of action) ?
Thank you!