Add feature/instructions to install from source
I was trying to access the new features from the latest PRs, and gave installing from source a try.
Attempting to install from the .git causes an error:
pip install "git+https://github.com/sign-language-processing/pose.git"
Collecting git+https://github.com/sign-language-processing/pose.git
Cloning https://github.com/sign-language-processing/pose.git to /tmp/pip-req-build-9auytf2j
Running command git clone --filter=blob:none --quiet https://github.com/sign-language-processing/pose.git /tmp/pip-req-build-9auytf2j
Resolved https://github.com/sign-language-processing/pose.git to commit d384d887ff4abaecec7f70fce9ad11b6a05aab94
Running command git submodule update --init --recursive -q
ERROR: git+https://github.com/sign-language-processing/pose.git does not appear to be a Python project: neither 'setup.py' nor 'pyproject.toml' found.
It is instead necessary to do something like:
conda create -n pose python=3.10 # otherwise you get latest python, which means no mediapipe!
pip install git+https://github.com/sign-language-processing/pose.git@master#subdirectory=src/python vidgear mediapipe psutil
# then you get the latest version of videos_to_poses
It seems like it would be nice to have this in the README, or a better method if there is one.
Note that the command above did not seem to work. I got errors with the pose_pipelines submodule. What does work is manually cloning, navigating to the pose/src/python subdir, and then doing pip install -e .
hmmmm what i do is:
git clone https://github.com/sign-language-processing/pose.git
pip install ./pose/src/python
I don't know if there's a way to do it with a single command.
Update: Tried a few things while trying to solve #174, no luck with the source install. https://colab.research.google.com/drive/1Dy9hy5NcF1cL6ndyQj-9smhxbAanbJ46#scrollTo=exVCYF3YWSeE
Seems like manually cloning and installing is still the way to go for now.