[Neurips'24] PyTorch Implementation of "MGF: Mixed Gaussian Flow for Diverse Trajectory Prediction". (https://arxiv.org/abs/2402.12238)

conda create -n mgf python=3.9
conda activate mgf
pip install -r requirements.txt

python src/data/TP/process_data.py

python src/test.py --scene {scene_name}

• scene_name = eth/hotel/univ/zara1/zara2/sdd

python src/train.py --model_name {model_name} --scene {scene_name} --gpu {gpu_id}

• scene_name = eth/hotel/univ/zara1/zara2/sdd
• Note: The real parameters are set in the config file, not in args.

• The basic code framework is based on FlowChain.