TensorFlow Lite module for MicroPython ?
Posted: Sat Jan 25, 2020 8:08 pm
Anybody working on making a "generic" TensorFlow Lite module for MicroPython ?
Maybe using the new v1.12 feature, dynamic native modules (in C), to avoid recompiling MicroPython firmware.
TensorFlow Lite is already on OpenMV M7/H7 and Sipeed MAix BiT running MicroPython.
Outside MicroPython, it is available for Apollo3 Blue/Artemis boards (with Cortex M4F @ 48/96 MHz, 384 kB RAM, 1 MB flash). There is also a community port for Teensy 3.x/4.0.
So my initial question is about make a TensorFlow Lite module compatible with more MicroPython boards, like Pyboard v1.1, Pyboard D, ESP32 (PSRAM only ?), etc.
TensorFlow Lite on MicroPython would be useful to analyse different sensor data, not only camera images, e. g., voice/sound from microphone, gestures from IMU (accelerometer, etc), spectra from nano-micro spectrometeres, etc.
Maybe using the new v1.12 feature, dynamic native modules (in C), to avoid recompiling MicroPython firmware.
TensorFlow Lite is already on OpenMV M7/H7 and Sipeed MAix BiT running MicroPython.
Outside MicroPython, it is available for Apollo3 Blue/Artemis boards (with Cortex M4F @ 48/96 MHz, 384 kB RAM, 1 MB flash). There is also a community port for Teensy 3.x/4.0.
So my initial question is about make a TensorFlow Lite module compatible with more MicroPython boards, like Pyboard v1.1, Pyboard D, ESP32 (PSRAM only ?), etc.
TensorFlow Lite on MicroPython would be useful to analyse different sensor data, not only camera images, e. g., voice/sound from microphone, gestures from IMU (accelerometer, etc), spectra from nano-micro spectrometeres, etc.