Skip to content

Latest commit

 

History

History

pix2pix

Folders and files

NameName
Last commit message
Last commit date

parent directory

..
tools
tools
 
 
README.md
README.md
 
 
pix2pix.py
pix2pix.py
 
 

README.md

Pix2Pix

This is a list of some tools to edit video, train pix2pix and export the model to be used in the pose2pose app.

Resize video to 512x256

ffmpeg -i input.mp4 -vf scale=512:256 -c:a copy output.mp4

Split video in frames

rm -r frames/*
ffmpeg -i input.mp4 -f image2 frames/img-%04d.png

Split in train and test

python3 tools/split.py --dir frames

Train

python3 pix2pix.py --mode train --output_dir train --max_epochs 200 --input_dir frames/train --which_direction BtoA

Test

rm -r test/*
python3 pix2pix.py --mode test --output_dir test --input_dir frames/val --checkpoint train

Export model

python3 pix2pix.py --mode export --output_dir export --checkpoint train --which_direction BtoA

Convert model to tensorflow.js

python3 tools/export-checkpoint.py --checkpoint export --output_file models/david_BtoA.pict

MISC

Cut a video

ffmpeg -i fornite.mp4 -ss 121 -t 29 -c:a copy fortnite2.mp4

Crop a video

ffmpeg -i fortnite1.mp4 -filter:v "crop=1280:690:0:30" -c:a copy fortnite1_crop.mp4

References

https://github.com/affinelayer/pix2pix-tensorflow

https://github.com/yining1023/pix2pix_tensorflowjs