Pytorch Implementation of TIM

This code is a Pytorch implementation of TIM. For implementation, I refer to the other matlab code. Original paper is Towards a practical lipreading system.

For use of TIM, please follow below steps.

1. Install required packages.

   pip install einops torch
   

2. Copy tim.py or clone this repository to your local machine.

Example

import torch
from tim import FixSequenceTIM

sample_video = torch.randn(18, 3, 224, 224)
tim_transform = FixSequenceTIM(sequence_length=16)
sample_video_transformed = tim_transform(sample_video) # shape: (16, 3, 224, 224)