search

AlexeyAB / darknet

YOLOv4 / Scaled-YOLOv4 / YOLO - Neural Networks for Object Detection (Windows and Linux version of Darknet )
http://pjreddie.com/darknet/
Other
21.75k stars 7.96k forks source link

What will happen if the xx.data files indicates wrong classes num? #3929

Open yiwc opened 5 years ago

yiwc commented 5 years ago

I have a training datasets of classes=3, but with a xx.data file wrote like this classes= 2 train = data/train_3.txt valid = data/test.txt names = data/gripper_3.names backup = backup/ cfg file is all right with classes=3 and other configurations are fine. What will will happen to the training result?

My wrong result weights seems very hard to give a right recognization of objects. Do you think it is caused by indicates wrong num in xx.data?

And why there is no warning since there is a classes num mismatch between xx.data file and train data img labeling txt or xx.cfg.

yiwc commented 5 years ago

I'm now trying updating my xx.data file into classes=3, and i'll see the result tomorrow. Thanks for your attention.

HagegeR commented 5 years ago

look at the readme, class and filter number should be changed according to the task

yiwc commented 5 years ago

look at the readme, class and filter number should be changed according to the task

Thank you! I mean the only difference is that I used different xx.data file, and the others files xx.cfg,xx.names is all set well as readme. I'll check new result (with right xx.data file) today to see what's going.