This is a python implementation of downloading and extracting the videos for mannequin paper "Learning the Depths of Moving People by Watching Frozen People."
- download the mannequin dataset from here
- extract the files in a directory (e.g. ~/MannequinChallenge)
-
I use pytube3 for downloading the videos from youtube. more information at here
-
Install pytube3 for python3:
pip install pytube3
-
make a folder code in the extracted directory each folder (train,test, and validation) and put the download.py there (e.g. ~/MannequinChallenge/train/code/download.py)
-
run the download.py:
python download.py
- your data will be downloaded in data folder in extracted directory (e.g. ~/MannequinChallenge/train/data/) each folder in the data directory represent each txt file.
-
once you downloaded the videos from youtube, you should extract the frames for each video based on the frame timesteps given in the txt files.
-
for that I use opencv library. you can install it:
pip install opencv-python
-
again for each dataset (train,test, or validation) put the extract.py in the code folder (e.g. ~/MannequinChallenge/train/code/extract.py)
-
run extract.py:
python extract.py