Darknet is an open source neural network framework written in C and CUDA. It is fast, easy to install, and supports CPU and GPU computation.
For more information see the Darknet project website.
For questions or issues please use the Google Group.
mkdir -p voc_training/backup
cd voc_training
wget https://pjreddie.com/media/files/VOCtrainval_11-May-2012.tar
wget https://pjreddie.com/media/files/VOCtrainval_06-Nov-2007.tar
wget https://pjreddie.com/media/files/VOCtest_06-Nov-2007.tar
tar xf VOCtrainval_11-May-2012.tar
tar xf VOCtrainval_06-Nov-2007.tar
tar xf VOCtest_06-Nov-2007.tar
python ../scripts/voc_label.py
wget https://pjreddie.com/media/files/yolov2.weights
../darknet partial ../cfg/yolov2-voc.cfg yolov2.weights yolov2.conv.20 20
python ../scripts/gen_anchors.py -filelist train.txt -output_dir generated_anchors/voc -num_clusters 5
cd ..
cp cfg/yolov2-voc.cfg cfg/yolov2-voc-test.cfg
Make sure the bath size and subdivisions are set properly in cfg/yolov2-voc.cfg
Training
batch=64
subdivisions=8
Copy the anchors generated in the last steps and replace the anchors in cfg/yolov2-voc.cfg and cfg/yolov2-voc-test.cfg
anchors = copy from "generated_anchors/voc/anchors5.txt"
Make sure the path are correctly set in the cfg/voc.data
classes= 20
train = voc_training/train.txt
valid = voc_training/2007_test.txt
names = data/voc.names
backup = voc_training/backup
mkdir voc_training/log
./darknet detector train cfg/voc.data cfg/yolov2-voc.cfg voc_training/yolov2.conv.20 >> voc_training/log/voc.log