Does not give the exact same result as darknet

[pfeatherstone]

Using the exact same image size (416x416), confidence threshold, nms threshold, config file (yolov3.cfg) and weights file (yolov3.weights), inferring on dog.jpg gives different results when using this repo and darknet.

(This is also the case with most implementations of yolov3 in other neural network frameworks (pytorch, keras). The DNN module in OpenCV is the only one that gives exactly the same results as darknet)

[pppplin]

you may want to checkout #255 , also noobj is so large in the default setting, see #295