This script downloads darknet and trains for one particular class. This script is only compatible with machines that have Nvidia drivers and CUDNN installed.
git clone https://github.com/sachinruk/YoloV3_train.git && cd ./YoloV3_train
- Change the password in
run.sh
file to something secure. - Run
bash run.sh
and go to localhost:8888 on your browser. You may need to do port forwarding to your local machine if you are running on a remote machine. In my case when connecting to remote instance I had to dossh -i aws_permissions.pem -L 8001:localhost:8888 ubuntu@$instanceIp
. - Enter the password that is stated on the
run.sh
file (please change the default) and open up a terminal in jupyter notebooks. - Run
bash train.sh -o snowman
for a snowman detector. All valid object types are listed undercategories.txt
.
Much of this code has been adapted from the blog post from learnopencv: https://www.learnopencv.com/training-yolov3-deep-learning-based-custom-object-detector/