Coral Edu projects for Raspberry Pi with simplified vision API

Clone this repo:


  1. 1bb3fe3 pure format cleanup by Scott Main · 24 hours ago main
  2. 606c1bb remove make_interpreter() in favor of pycoral version by Scott Main · 2 days ago
  3. 7cc0535 Refactor the code as a proper Python package. by Scott Main · 3 days ago
  4. bc0c9c7 refactor module name from aiy.coral to coralkit by Scott Main · 3 days ago
  5. c3b12eb fix heading levels in api reference by Scott Main · 3 days ago

Coral Kit API and examples

The coralkit Python APIs in this repo greatly simplify the amount of code required to perform common operations with TensorFlow Lite models, such as running inference with image classification, object detection, and pose estimation models. It also includes scripts to collect training images and perform transfer learning with an image classification model.

This project is designed specifically for a Raspberry Pi with a Coral USB Accelerator and a camera (though it may be repurposed for other systems as well).

For detailed documentation using the Raspberry Pi, including the coralkit API reference, see


If you're on a Raspberry Pi, follow the setup guide at

For other manual installs, you must first install the libedgetpu and pycoral libraries. And we assume you are using the Coral USB Accelerator, so follow the instructions here.

Then clone this repo and install as follows:

git clone

cd coralkit

python3 -m pip install .