HOG vs YOLO Detector - OpenPTrack/open_ptrack_v2 GitHub Wiki

While Histogram of Oriented Gradients(HOG) had previously been shown to be a useful and widely utilized approach to object detection, with the rise of convoluted neural networks and deep learning, we now have better methods for object detection.

One of such methods is YOLO (You Only Look Once) which is based on deep learning. OpenPTrack v2 "Gnocchi release" uses YOLO v2 and there are plans to implement YOLO v3 in the future.

Compared to HOG, YOLO is more accurate and much faster. However, YOLO requires more computational power.