README.md

Code for "Distilling the Unknown to Unveil Certainty"

##Standard Network Create a folder named "checkpoints," and download the standard network from the link below. Place it inside this folder.

• The standard network learned from CIFAR10
• The standard network learned from ImageNet

Datasets

In-distribution Datasets

• CIFAR10
• Imagenet

Out-of-Distribtion Datasets

Create a folder named "data," and download the OOD datasets from the link below. Place it inside this folder.

• CUB200
• StanfordDogs120
• OxfordPets37
• Oxfordflowers102
• Caltech256
• DTD47
• COCO
• Please refer to OpenOOD for more OOD benchmarks.

Run CA-

python Generator.py --T 1000 --gap 200 --lr 0.05 --save --generator 'DI'
python Discriminator.py --epoch 200 --T 1000 --gap 200 --generator 'DI'

Run CA+

python Generator.py --T 1000 --gap 200 --lr 0.05 --save --generator 'DR'
python Discriminator.py --epoch 200 --T 1000 --gap 200 --generator 'DR'