This repository provides several synthetic near-OOD datasets for ImageNet-1k,
as introduced in the ICCV 2025 paper (highlight) SynOOD.
You can also generate these datasets yourself using our official code repository.
Each dataset contains 50 synthetic near-OOD training images for each of the 1,000 ImageNet-1k categories.
These images are highly similar to the in-distribution samples of their corresponding categories, but are
located near the decision boundary… See the full description on the dataset page:
https://huggingface.co/datasets/Grey0423/nearOOD_of_SynOOD.
