Docker for mmhuman3d - open-mmlab exposing streamlit a app.
mmhuman3d is an Open Source project PyTorch-based the use of 3D human parametric models, it is a part of the OpenMMLab project.
mmhuman3d is a 3D human mesh recovery toolkit which enables the generation of 3D human models from 2D images or videos.
The container can be downloaded at or it can modified editing the file docker/Dockerfile.
mmhuman3d needs GPU for inference, so you need to support GPU inside your container.
It depends on the image nvcr.io/nvidia/pytorch:22.08-py3. You need to create an account at create NGC and generate an api key.
Login into ngc:
docker login nvcr.io
Username: $oauthtoken
password: [insert here your api key]
mkdir vis_results
mkdir videos
docker run --gpus all --ipc=host --ulimit memlock=-1 --ulimit stack=67108864 -it --rm -v $(pwd)/videos:/app/mmhuman3d/videos -v $(pwd)/vis_results/vis_results:/app/mmhuman3d/vis_results -p 8501:8501 nvcr.io/nvidia/pytorch:22.08-py3_mmhuman3d
Open the browser at 'http://localhost:8501'