1050-105 is our 2021 Vision system which combines the benefits of a traditional OpenCV system with that of a Machine Learning model. We use traditioanl OpenCV filtering and masking tecniques alongside a TensorFlow Lite Object Detection model to accurately detect and pinpoint the location of a vision target or gamepiece.
or in meme form...
- COCO dataset: Roboflow
- TFRecord dataset: Roboflow
- Train SSDMobileNet v2 model: Google Colab NB
- Convert to TFLite: Google Colab NB