VFP_ODE

Project for Image and Video Processing class at NYU Tandon.

To set up the environment for the project:

1. Create a conda environment:

conda create --name ivp 

2. Activate the environment:

conda activate ivp

3. Install the dependancies:

pip install -r torchdiffeq/requirements.txt

To run the code:

Steps to train the model:

1. Enter in the project directory:

cd torchdiffeq

2. To train a model which uses ODE solver to model the intermediate hidden states:

python conv_gru_with_mask.py

Alternatively, if running on a slurm based backend, clone the slurm branch and use:

sbatch convtrain.sbatch

3. To train a model with vanilla ConvGRU layers:

python conv_gru.py

4. To validate the model, set the --val argument to True :

python conv_gru_with_mask.py --val=True