search

AlexeyAB / darknet

YOLOv4 / Scaled-YOLOv4 / YOLO - Neural Networks for Object Detection (Windows and Linux version of Darknet )
http://pjreddie.com/darknet/
Other
21.65k stars 7.96k forks source link

Overfitting Yolov2 question #160

Open smallesan opened 7 years ago

smallesan commented 7 years ago

I am trying to overfit the Yolov2 for specific items (50 in total). I have had some amount of success by training for an extended time, but is there any parameters that I can tune to make it overfit even better? Thanks.

MyVanitar commented 7 years ago

increase the learning rate, use smaller dataset

MyVanitar commented 7 years ago

Also please tell us why do you want to overfit the YOLO. Overfitting is a nightmare in deep learning models. is there any reason?

gizmoy commented 6 years ago

Yup, there is. Overfitting small portion of your dataset is a sanity check before you start 'heavy' training. If you are not able to overfit 10 examples something is probably wrong.

bit-scientist commented 6 years ago

does anybody know how to overfit the yolov2 from scratch just to prove sanity checks are really helpful? Any links, posts are appreciated.