This is a drawing detection application for raspberry-pi. It detects if a drawing is a car or not. There are two modes of running:

• take a picture of your drawing
• draw directly on the tactile screen of the raspberry

Create a python3 venv and all the required packages and the app in your virtual env. At the root of the project, run:

python3 -m venv venv
pip install -r requirements
pip install -e .

pytest

Run the folllowing command to load the app with camera:

car_detector camera

With clicking mouse drawing and optionnal arguments:

car_detector draw --model <path/to/model> --saved <path/to/save/image>

The modelisation notebook can be found at the root of the project. The data comes from the google quick draw application. Usefull command for port forwarding (if you use ec2 instance):

ssh -L <local:port>:localhost:<distant:port> <user>@<ip-adress>
# example to forward tensorboard:
ssh -L 8881:localhost:6006 [email protected]

Article 1: train your model

Article 2: deploy it on a raspberry

Article 3: use the model and enhance it

### Schemes

#### Use Ansible to configure Raspbian on QEMU

#### Pipeline overview