Mingze Xu

You just need to convert the feature of each video from .pkl into a single .npy file. The feature should be in shape of L x C, where L is...

1) Download the dataset from HRI; 2) prepare the data structure as in README's data preparation; 3) add HDD to "data/data_info.json" (you can refer https://github.com/xumingze0308/TRN.pytorch/blob/master/data/data_info.json).

LSTR and TRN use the same data structure, so maybe you can directly use what you have. I don't have the inputs for HDD. They are owned and maintained by...

Each numpy file contains the feature for one particular video and is in shape of "L x C", where L is the number of frames and C the feature dimension.

Sorry, I don't think we have that plan.

PRED_SCORES_FILE is the output result after you run test_net.py, which stores the action detection scores of each frame in each video. eval_perstage.py will evaluate the model in other metrics based...

L is the number of frames in a particular video. 22 is the number of classes, where the value is either 0 or 1 --- 0 means this action doesn't...