Darknet: Train YOLO using own data set

[rucsacman]

I want retrain the yolo for only detect single class as a exmaple only want detect "STOP" sign .So i want remove all other weight files and I want only my weight file how can i do this.

[ahsan856jalal]

change the yolo-voc.cfg file ,batches=64,subdivision=8, last convolution layer's filter=30 and classes in region layer to be 1 and modify your_data.data and your_data.names files . then start training
./darknet detector train cfg/your_data.data cfg/yolo-voc.cfg darknet19_448.conv.23
Then use this trained model for one class detector

[rucsacman]

Can you explain bit more IM very new to YOLO

[ghost]

I made a document to help me and my colleagues remember how to do this. Maybe it could help you too: https://github.com/drforester/Guide_to_using_YOLOv2

[rucsacman]

Thank you !!

On Tue, Feb 27, 2018 at 8:15 AM, David Forester [e-mail]
wrote:

I made a document to help me and my colleagues remember how to do this.
Maybe it could help you too: https://github.com/drforester/
Guide_to_using_YOLOv2

—
You are receiving this because you authored the thread.
Reply to this email directly, view it on GitHub
https://github.com/pjreddie/darknet/issues/487#issuecomment-368728088,
or mute the thread
https://github.com/notifications/unsubscribe-auth/Aav-OTzoKI7_auIdlVwV8sLNj79xIOaKks5tY2xhgaJpZM4SQrXg
.

[mehditlili]

@rucsacman did you fork the repository containing the guide? drforester account has been deleted

[t-mullen]

It was archived: https://web.archive.org/web/20180617164046/https://github.com/drforester/Guide_to_using_YOLOv2