/Speech_project_Vin

Dataset

We using dataset CREMA-D Link Dataset

NOTE

Pretrained Weight for Multimodal

  1. Audio_Encoder
  2. Visual_Encoder

Training Guidline This source code train with Kaggle using Pytorch version

  1. You need pip install timm version 0.4.5 for load Audio_Encoder
  2. Using config.yaml to fix the name folder dataset, weight load
  3. Run train_kaggle.inypb for training