- Install opencv on your system for the Python bindings to work
sudo apt install libopencv-dev - Create and activate virtual Python environment if you like (e.g.
python3 -m venv .venv, followed bysource .venv/bin/activate) - Install dependencies
pip install -r requirements.txt
-
Download EfficientDet model of desired size from the TF2 detection zoo
https://github.com/tensorflow/models/blob/master/research/object_detection/g3doc/tf2_detection_zoo.md -
Extract .tar.gz file
tar xvf efficientdet_d*.tar.gz
- Put some images into
./imagesor somewhere else (if you choose a custom directory you have to supply it to the script below) - Run
object_detect.py(Run it with--helpto get some usage hints) - Find your annotated images in
./annotated_images(by default)