This is a docker built for openpose https://github.com/CMU-Perceptual-Computing-Lab/openpose
216
The docker is built from nvidia/cuda:11.7.1-cudnn8-devel-ubuntu20.04 for OpenPose software from CMU. Note that OpenPose does not support cudnn8, so the OpenPose inside the docker is not built with cudnn to avoid memory problems, see issue.
First, use this guide to install nvidia-docker
Then, pull the image:
docker pull hychiang/openpose
To run the container, use the following command (Ubuntu):
$ xhost + # optional, if you want to display
$ docker run -it -e DISPLAY=$DISPLAY -v /tmp/.X11-unix:/tmp/.X11-unix --gpus all hychiang/openpose /bin/bash
We need to -v /path/to/images:/images if we want to process any images using openpose.
-e DISPLAY=$DISPLAY -v /tmp/.X11-unix:/tmp/.X11-unix flags are optional. Use the flags if you want to display.
Processing images:
# host
$ xhost + # optional, if you want to display
$ docker run -it -e DISPLAY=$DISPLAY --gpus all -v /tmp/.X11-unix:/tmp/.X11-unix -v /path/to/images:/image hychiang/openpose /bin/bash
# inside docker
$ cd /openpose
$ ./build/examples/openpose/openpose.bin --image_dir /images --display 0 --write_json /output --write_images /output_rgb
or
Only running inference with the docker environment
# host
$ docker run --rm --gpus all -v /path/to/images:/images:/images --workdir /openpose hychiang/openpose ./build/examples/openpose/openpose.bin --image_dir /images --display 0 --write_json /output --write_images /output_rgb
Content type
Image
Digest
sha256:db5072c34…
Size
4.9 GB
Last updated
almost 4 years ago
docker pull hychiang/openpose