GOLOv2

YOLOv2 for Golang

This projects implements the yolov2 (https://pjreddie /darknet/yolov2/) RegionLayer in Go. It is heavily inspired by duangenquan's C++-RegionLayer implementation (https://github /duangenquan/YoloV2NCS).

This projects makes use ofgocv(https://gocv.io) andgo-ncs(https://github /hybridgroup/go-ncs/), both from hybridgroup (https://github /hybridgroup).

It comes with atiny-yolo caffe modelwhich I derived from original weights (https://pjreddie /media/files/yolov2-tiny-voc.weights) with thisdarknet2caffe converter:https://github /marvis/pytorch-caffe-darknet-convert.It also comes with aMovidius NCS model version of tiny-yolowhich I compiled from the converted caffe model.

Setup

1. Installgocvas described onhttps://gocv.io/getting-started/
2. Installgo-ncsas described onhttps://github /hybridgroup/go-ncs
3. Plug in your Movidius Neural Compute Stick
4. $ git clone git@github:dymat/GOLOv2.git
5. $ cd GOLOv2
6. go run *.go