Motion_Dataset

# Motion Dataset Repository ## Overview This repository stores the motion datasets used in our Text-to-Motion (T2M) research pipeline. It contains both real human motion capture data and synthetic motions generated by diffusion models, along with their processed representations and rendered animations. The repository mainly includes: - Raw motion files from AMASS - Processed HumanML3D-style 201D motion representation - Motion data generated by the Hunyuan Motion model - Rendered Blender animations for visualization and video-based experiments These datasets are used for: - Text-to-Motion training - Motion captioning / video understanding - Motion representation research - Synthetic motion generation evaluation