Chalearn LAP Inpainting Competition Track 2 - Video decaptioning

This is repository for Track2 of Chalearn LAP Inpainting Competition - Video decaptioning

Instruction for Training

• Firstly, convert all your video files (video with subtitles and without subtitles) into tensorflow records in folder named train_records in source folder by running preprocess_train_images.py.
• Once tf records are made, you can start training by running main.py
• You can also specify attributes according to hardware i.e. python main.py --use_tfrecords True --batch_size 16
• You can edit model architecture in model.py

Instruction for Testing

• Once model is trained and weight files are updated you can call test.py to see outcome of trained model.
• Specify source destination of test videos(with subtitles) in model.py (currently it is ~/Inpaintin/dev/X)
• Output videos will be stored in folder named out_video with name changed as 'X' will be replaced by 'Y' in name.

Sample result from testing