"Real-time" Hand-Tracking Pong. As of the moment, object detection inference time is the bottleneck, at 1-2 FPS on a Macbook Pro (i5, 2.5GHz, 8GB) with detections on 160x90 images.
The original hand tracker utilities and model are taken from Victor Dibia's Hand Tracking repository.
-
Setup a Python virtual environment with Python 3.5.
-
Install all Python dependencies.
pip install -r requirements.txt
For Ubuntu users, install
libsm6for OpenCV to work:apt install libsm6
Installation is complete. To get started, launch the demo. I recommend running the multi-threaded version.
python main_multi.py