Skip to content

Latest commit

 

History

History
13 lines (10 loc) · 733 Bytes

dataset_curation.md

File metadata and controls

13 lines (10 loc) · 733 Bytes

How to structure your files before running extract_videos.py

  • Place extract_videos.py in the same directory (now referred to as "root") as "hmdb51_org" and the splits folder
  • Rename folder containing splits text files to "splits"
  • Make a folder named "data" in the root directory
  • Make folders named "test" and "train" in the "data" folder
  • Make sure hmdb51 dataset folder is named "hmdb51_org" and that everything inside is extracted
  • Run extract_videos.py to completion - should see action folders appearing in "test" and "train"

How to get dataset batch files

  • Define data folder and create folder within named "dataset_{}" [number of frames] and folders within named "test" and "train
  • Run dataset.py