Code to run experiments for my OOD thesis project.

Some OOD detection methods code is adapted from other sources:

• Odin, MSP and Energy were adapted from: https://github.com/deeplearning-wisc/large_scale_ood [1]
• DDU was adapted from https://github.com/omegafragger/DDU/ [2]

Dataset downloads:

• Cats and dogs: https://www.robots.ox.ac.uk/~vgg/data/pets/
• Imagenet validation: https://www.kaggle.com/c/imagenet-object-localization-challenge/overview/description
• Mammals: https://www.kaggle.com/datasets/asaniczka/mammals-image-classification-dataset-45-animals
• Urban-100: https://paperswithcode.com/dataset/urban100
• Cars and Busses: https://universe.roboflow.com/deeplearning-lryeh/car-bus-dataset
• Textures: https://www.robots.ox.ac.uk/~vgg/data/dtd/
• Celebrity faces: https://huggingface.co/datasets/tonyassi/celebrity-1000
• Animal faces: https://www.kaggle.com/datasets/andrewmvd/animal-faces

[1] Huang, R., & Li, Y. (2021). MOS: towards scaling out-of-distribution detection for large semantic space. CoRR, abs/2105.01879. https://arxiv.org/abs/2105.01879

[2] Mukhoti, J., Kirsch, A., van Amersfoort, J., Torr, P. H. S., & Gal, Y. (2021). Deterministic neural networks with appropriate inductive biases capture epistemic and aleatoric uncertainty. CoRR, abs/2102.11582. https://arxiv.org/abs/2102.11582