Open mocleiri opened 2 years ago
I was watching one of the Shawn Hymel Tinyml videos and at that time tflm didn't have the right kernels imported to support object detection.
So this may involve porting the applicable ops from tflite.
Edge impulse has their FOMO model but that probably won't run without their custom libraries.
Recently tflite made a new example implemented using tflite python on raspberry pi.
https://github.com/tensorflow/examples/tree/master/lite/examples/object_detection/raspberry_pi
https://youtu.be/mNjXEybFn98
The tflite model size is 4MB so can be tried on Teensy 4.1 or other boards with psram enabled.
The first step would be to implement #10 so that our api is a subset of the tflite python api.