This is an experimental project, no support will be done.
Works only on Android for now.
It uses:
- Kivy android camera to grab RGBA pixels
- Pillow to convert the image to the tensor input format
- TensorFlow lite Android libraries to detect
This example is mostly based on the Image classification from TensorFlow lite example repository.
- Train your model https://teachablemachine.withgoogle.com/
- Export as a Quantized model for TensorFlow lite
- Copy it into the app, ajust the paths.