philipperemy / yolo-9000

YOLO9000: Better, Faster, Stronger - Real-Time Object Detection. 9000 classes!
Apache License 2.0
1.18k stars 309 forks source link

How to train yolo9000? #11

Closed xlxlcyRuSReXxlxl closed 6 years ago

xlxlcyRuSReXxlxl commented 6 years ago

Using the pre trained yolo9000 model with Python I notice that is much less accurate than yolov2, and I guess it's normal because it has to detect more classes. Right now I am looking for detecting landscapes (as sea shore, lake shore, mountains, deserts and things like these), so I was wondering if you could give me some guidelines to train the weights since I have no background in this. Beside the code I would like to know a good sample size for each class and the if it takes a long time with a fairly good machine. Thank you a lot !

philipperemy commented 6 years ago

@xlxlcyRuSReXxlxl that's a very good question for which I don't have the answer.

This project does not contain the source code of Yolo.

In your case, you should ask directly to the guys here: https://groups.google.com/forum/#!forum/darknet

Also if you find out how to do that, that would be interested to include it inside this repository!

Thanks!

xiaoxiongli commented 5 years ago

@xlxlcyRuSReXxlxl Do you successfully train yolo9000? can you give me some tips about how to train yolo9000?