Easy training on custom dataset. Various backends (MobileNet and SqueezeNet) supported. A YOLO demo to detect raccoon run entirely in brower is accessible at https://git.io/vF7vI (not on Windows).
MIT License
1.73k
stars
787
forks
source link
how much epoch need to get good accuracy on pascal dataset with 20 classes.? #380
@rodrigo2019 @hobson @mrshu @sirotenko @usatenko @wikiwen @BoshengW
I'm training yoloV2 model with Pascal, for 50 epoch it takes more than 6hr and give loss=2.36 and predict the same bounding box for every image.
so on how much epoch I have to train yolov2 to get good results and how to speed the training process.
rodrigo2019
commented
5 years ago
@rodrigo2019 @hobson @mrshu @sirotenko @usatenko @wikiwen @BoshengW I'm training yoloV2 model with Pascal, for 50 epoch it takes more than 6hr and give loss=2.36 and predict the same bounding box for every image. so on how much epoch I have to train yolov2 to get good results and how to speed the training process.