Played around with MNIST until i got >99.50% accuracy. The model I used is a beefier LeNet with Dropout and Batch Normalization. Rand Augment was used to combat overfitting and increase overall accuracy.
Neural network architecture (BeefyLeNet):
Procedurally-Generated-Human/MNIST-99.50
Folders and files
| Name | Name | Last commit date | ||
|---|---|---|---|---|