PyTorch implementation of Fast Neural Style Transfer (official Lua implementation).
I have made some trained models available here.
be sure that you have installed ffmpeg, and pkg-config correctly.
pip install av skvideo
-
apt-get install -y \ libavformat-dev libavcodec-dev libavdevice-dev \ libavutil-dev libswscale-dev libavresample-dev
python3 train.py --dataset_path <path-to-dataset> \
--style_image <path-to-style-image> \
--epochs 1 \
--batch_size 4 \
--image_size 256
Figure: Training progress over the first 10,000 batches.
python3 test_on_video.py --video_path <path-to-video> \
--checkpoint_model <path-to-checkpoint> \
python3 test_on_image.py --image_path <path-to-image> \
--checkpoint_model <path-to-checkpoint> \