AMDM:Animal Motion Diffusion Model

We train an action2motion model and a text+scene2motion model. For example, input action lable, the model generate SMAL motion sequence.

Generated Results

run

walk to the bed

More results in demo/.

Try Pretrained Models

👉 https://huggingface.co/Lebin749/Animal_Motion_Diffusion_Model

Environment

conda create -n animal python=3.12 -y
conda activate animal
pip install -r requirements.txt

Data Preparation

Prepare the required datasets under data/dataset/:

• action_motion_data/ for action-conditioned motion generation
• scene_text_motion_data/ for text + scene conditioned motion generation
• scannet++/data/ for scene data used by text-scene task

Training

python train.py

Inference

python infer.py