A look into the mislabelled objects in the COCO dataset with a focus on the traffic lights.
COCO dataset with refined traffic light labels red, green, and na for training driving-related object detectors and beyond.
How COCO annotations are structured and how to use them to train object detection models in Python.
How to estimate the distance to objects of known width using a single camera and an object detector in Python with OpenCV and Pytorch.
Lane line detection as example for the limits of non-learning based techniques.