Testing body-pix and tensorflow with Rust.
This is just a demo project to learn how to use Tensorflow with a frozen model.
Build the application and start it. Use the control panel to open and process images.
The build-script depends on python3 and the package tfjs_graph_converter. It
should be available through a simple installation with pip:
pip install tfjs_graph_converter
The build script will then download and convert the body-pix models
This should work on all major platforms, as long as you have Python 3 installed for the build. However, I have only tried it on Linux - so feedback and patches are welcome!
Sending me a message on GitHub is probably the best way to get in touch
A big shout-out to Ajai Chemmanam (ajaichemmanam on Github), and his simple_bodypix_python project. I owe my understanding of how to massage the input and output to improve prediction to this project. (Which I have been tinkering with a lot.) Check it out here: simple_bodypix_python