This is an code implementation of "Improving Transferability for Cross-domain Trajectory Prediction via Neural Stochastic Differential Equation", AAAI'24

Please follow below steps to run our code

1. Create virtual environment in Anaconda with env.yml

conda env create --file env.yaml -n trajsde
conda activate trajsde

2. Prepare raw dataset of nuScenes and Argoverse

Download meta data of trainval set of nuScenes from "https://www.nuscenes.org/nuscenes#download".

Download Training/Validataion/Testing dataset of motion forecasting from "https://www.argoverse.org/av1.html#download-link"

Locate them in 'data' dir as following:

.
├── configs
├── ...
├── data
│   ├── nuScenes
│   │   ├── maps 
│   │   ├── samples
│   │   ├── ...
│   │   └── v1.0-trainval
│   └── argodataset
│       ├── map_files
│       ├── train
│       └── val
└── train.py

3. Run preprocessing for nuScenes and Argoverse, respectively

mkdir preprocessed
# Argoverse
python dataset/Argoverse/Argoverse_abs.py
# nuScenes
python dataset/nuScenes/nuScenes_hivt.py

Then, preprocess data files are saved in 'preprocessed/Argoverse' for Argoverse and 'preprocessed/nuScenes' for nuScenes.

4. Make checkpoints dir and run training code

mkdir checkpoints
# Vanilla HiVT 
python train.py -n baseline -c configs/nusargo/hivt_nuSArgo_trmenc_mlpdec.yml
# Ours
python train.py -n nsde -c configs/nusargo/hivt_nuSArgo_sdesepenc_sdedec.yml