Add support for Yolo v3 tiny & Yolov3

[deimsdeutsch]

Kindly add support for Object Detection using:

https://github.com/AlexeyAB/darknet

It is very fast and accurate. Particularly Yolov3 and Yolov3-tiny.

[beniz]

Hi, thanks for the recommendation. Yolo is a good architecture indeed.

At the moment we support a range of embedded to desktop best practice architectures, i.e. from squeezenet-SSD, mobilenet-SSD to refinedet and detectron.

Our experience is that the architectures above span across multiple platforms well, from embedded to high end desktop GPUs.

Yolo-v3 has excellent capabilities, and at the moment we would favor integration via one of the large scope backends such as Caffe or Caffe2.

For references: https://github.com/dongfangduoshou123/Caffe2-yolo-v3 and https://github.com/dongfangduoshou123/Caffe2-yolo-v3/blob/master/demo.py

https://github.com/ChenYingpeng/caffe-yolov3