titu1994 / DenseNet

DenseNet implementation in Keras
MIT License
707 stars 294 forks source link

poor cifar100 results #39

Closed jcpeterson closed 5 years ago

jcpeterson commented 6 years ago

The standard cifar100 model only obtains about 60% validation accuracy when letting the standard script run for as many epochs as it wants before stopping early.

titu1994 commented 6 years ago

Interesting. I have tried Cifar 100 with this model a long time ago and it did get into the 70s at least. I didn't use early stopping though, and ran for 300 epochs, with learning rate halved when there was no improvement after 20 epochs.