non-maxima suppresion after feature generation is computationally wasteful

[anlutfi]

In object_tracker.py you first do detections = [Detection(bbox, score, class_name, feature) for bbox, score, class_name, feature in zip(converted_boxes, scores[0], names, features)] and then indices = preprocessing.non_max_suppression(boxs, classes, nms_max_overlap, scores)

I suggest that you perform NMS immediately after running yolo, so you don't waste time computing boxes that will just end up unused