caffe-yolov2-windows

Reference

https://github.com/eric612/Vehicle-Detection

https://github.com/eric612/MobileNet-SSD-windows

https://github.com/gklz1982/caffe-yolov2

https://github.com/duangenquan/YoloV2NCS

Modifications

  1. caffe training
  2. add pre-trained model
  3. fix bugs
  4. windows support
  5. vehicle detection

Configuring and Building Caffe

Requirements

  • Visual Studio 2013 or 2015
  • CMake 3.4 or higher (Visual Studio and Ninja generators are supported)
  • Anaconda

The build step was the same as MobileNet-SSD-windows

> cd $caffe_root/script
> build_win.cmd 

GPU Building

config build_win.cmd "CPU_ONLY" = 0

Usage

convert model

cd examples

  1. convert yolo.cfg to yolo.prototxt
  2. convert yolo weights to caffemodel

Detection

cd caffe_root

  1. Download pre-trained caffmodel , save at location $caffe_root/models/yolo/
  2. examples\yolo_detection.cmd

If load success , you can see the image window like this

alt tag

Trainning

There has two ways for training your dataset

  1. training with caffe (recommand)
  2. training with darknet project and convert the weights to caffemodel

training with caffe

Prepare data option 1(recommand)

Save lmdb at location $caffe_root\data\yolo\

Prepare data option 2

Put your dataset into $caffe_root\data\yolo\VOCdevkit

> cd $caffe_root\data\yolo
> python get_list.py 

Check files label_map.txt,trainval.txt,test_2007.txt are all correct

> cd $caffe_root\data\yolo
> create_dataset.cmd

Run training

> cd $caffe_root
> examples\yolo_train.cmd

Training with darknet project

yolo-windows

yolo-windows-my-version

Note

  1. There has no data augmentation code inside (ex.hue,rotation) , note the training mAp was bad
  2. I still try to train a good caffemodel , currently ,the deploy model was just show how to do training

Future

  1. Add MobileNet-YOLO
  2. Data augmentation