Video-Colorization

Link to the presentation

Steps to train the models

• Make a folder train_videos in the root directory of the repo and put any number of 640 X 360 color videos in the directory.
• Run python3 scripts/frame_extract.py to extract the color frames. It is important that the command is run from root directory.
• Run python3 train_scripts/1_basic.py for training the basic model. The model would be saved in models/1_basic.h5 after each epoch. Again, it is important for the commands to be executed from root folder.
• Similarly run python3 train_scripts/1_transfer.py and python3 train_scripts/1_transfer2.py and run models with transfer learning.

Steps to test

• Put a 640 X 360 video inside test_videos named testvid1.mp4. Ultimately, the file should be test_videos/testvid1.mp4.
• To run prediction of basic model, run python3 test_scripts/test.py models/1_basic.h5. The predictions would be created in models/1_basic.h5d/.
• To run the prediction of other models, run python3 test_scripts/test_final.py models.1_transfer*.h5. The predictions would be created in models/1_transfer*.hd5/