[NeurIPS 2024] E-Motion: Future Motion Simulation via Event Sequence Diffusion

Arvix: E-Motion: Future Motion Simulation via Event Sequence Diffusion.

🛠️ Requirements and Installation

• Python >= 3.10
• Pytorch == 2.0.1
• CUDA Version >= 11.7
• Install required packages:

git clone https://github.com/p4r4mount/E-Motion.git
cd E-Motion
conda env create -f environment.yml

📜 Datasets Preparation

Download the training data from EventVOT, and process the data with following command:

python utils/DataPreprocess.py --dataset_dir /path/to/dataset

🚅 Training

accelerate launch train.py \ 
	--num_processes num_processes \
	--main_process_port main_process_port \
	--config /path/to/config

💡 Pre-trained weights

Google Drive: Checkpoint

🚀 Inference

Download some samples from Google Drive，and run the following command for inference:

python predict.py --model_path /path/to/model/checkpoint \
                  --data_path /path/to/data/file.npy \
                  --output_path /path/to/output/directory

Citation

@article{wu2024motion,
  title={E-Motion: Future Motion Simulation via Event Sequence Diffusion},
  author={Wu, Song and Zhu, Zhiyu and Hou, Junhui and Shi, Guangming and Wu, Jinjian},
  journal={Advances in Neural Information Processing Systems},
  volume={37},
  pages={105552--105582},
  year={2024}
}