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
784
forks
source link
How to compute accuracy metric during training and not only loss and val_loss?? #280