Vikr-182 / ddn-forecasting

Trajectory Forecasting using Deep Declarative Networks

#DDN Forecasting

Installation

git clone https://github.com/Vikr-182/ddn-forecasting

Dataset

Find the top 1000sq in data/ folder.

Training

python3 script_train.py --train_dir <path/to/train/dataset> --test_dir <path/to/val/dir> --lr 0.0005 --traj_save_path <path/to/saved/results> --end_epoch 10

If using the MLP model, make sure to flatten using the --flatten option -

python3 script_train.py --train_dir <path/to/train/dataset> --test_dir <path/to/val/dir> --lr 0.0005 --traj_save_path <path/to/saved/results> --flatten

Test

python3 script_train.py --test --train_dir <path/to/train/dataset> --test_dir <path/to/test/dir> --traj_save_path <path/to/saved/results>