@Jingkang50 Here are all of the remaining commits from my side. The detailed change log can be found in the wiki. Notable outcomes include:
More comprehensive results on CIFAR-10, CIFAR-100, ImageNet-200 (OOD and FS-OOD), and ImageNet-1K (OOD and FS-OOD). For ImageNet-1K we have results on ResNet50, Swin-T, and ViT-B-16.
An easy-to-use evaluator where people only need to provide a pre-trained classifier, specify the corresponding ID dataset, and specify a postprocessor. There is a notebook tutorial example_imagenet_eval.ipynb on this, allowing people who are unfamiliar with OOD detection to test their models with minimum effort.
OpenOOD can now be installed by pip install git+https://github.com/Jingkang50/OpenOOD.git
I expect some updates will be needed for the Readme later.
@Jingkang50 Here are all of the remaining commits from my side. The detailed change log can be found in the wiki. Notable outcomes include:
example_imagenet_eval.ipynb
on this, allowing people who are unfamiliar with OOD detection to test their models with minimum effort.pip install git+https://github.com/Jingkang50/OpenOOD.git
I expect some updates will be needed for the Readme later.