search

hbdat / eccv20_Multi_Task_Procedure_Learning

Self-Supervised Multi-Task Procedure Learning from Instructional Videos @ ECCV20
MIT License
6 stars 2 forks source link

Possibly incorrect frame count in class ProceLDataset #4

Open Sid2697 opened 3 years ago

Sid2697 commented 3 years ago

Hey!

While extracting features for the ProceL dataset, I noticed that you have added the self.num_files by 1. https://github.com/hbdat/eccv20_Multi_Task_Procedure_Learning/blob/7783e7b9b47498933d359adfd761579f09934c3c/core/ProceLDataset.py#L36

Due to the addition, the rest of the code was taking one extra file and raising an error when the corresponding image file was not present. However, I was able to run through all the frames without adding 1. Could you please let me know why you had to do it?

Also, here: https://github.com/hbdat/eccv20_Multi_Task_Procedure_Learning/blob/7783e7b9b47498933d359adfd761579f09934c3c/core/ProceLDataset.py#L38 for me {0:0>7} works as there were 7 digit long image names. I am not sure at what stage the naming convention changed. Do let me know if you have any idea about it.

Regards, Siddhant Bansal