YOLOv3 support

[dkurt]

This pullrequest changes

YOLOv3 support

resolves #11310

Efficiency measurements:

Backend, Target	Median time per 416x416 image
DNN_BACKEND_DEFAULT, DNN_TARGET_CPU	216.11ms
DNN_BACKEND_DEFAULT, DNN_TARGET_OPENCL	353.77ms

CPU: Intel® Core™ i7-6700K CPU @ 4.00GHz x 8
GPU: Intel® HD Graphics 530 (Skylake GT2)

Merge with extra: opencv/opencv_extra#459

[vpisarev]

another great new feature! 👍

[SoullessStone]

Thanks a lot!

[ASShusharin]

Hi, I try to run dnn/object_detection.cpp with params: -c=yolov3.cfg -m=yolov3.weights -classes=coco.names --scale=0.00392 --width=416 --height=416 and it works, but time it takes for one image is about 4 seconds. Function: net.forward(outs, getOutputsNames(net)) takes the most time.
Although on the image (https://drive.google.com/open?id=1kZIUS5Dct3sPRZrMKsixFCWLgizbKnzd) on top is written about 745 ms. I run in release and use opencv4.
Also on top of this page is written that median time per 416x416 image is 216.11ms.
What may be the reason of this time difference?
I have Ubuntu OS: https://drive.google.com/open?id=1MvbLah4gNFB-P2JvW5P5MZ7znhlX5u8Y