/WaterCam

a RPi camera for on-device segmentation of water

Primary LanguagePythonMIT LicenseMIT

WaterCam

Raspberry Pi Python TensorFlow

A camera for segmenting water images in the field.

The device is a Raspberry Pi 4, Hyperpixel display, Camera Module 3, and an on-board ML model that was made with the RIWA dataset and Segmentation Gym.

Using a Google Coral edgeTPU, a 256 x 256 model can run at 3 frames/second. Without an accelerator, pictures takes ~3 seconds to segment.