OmniLens releases the AODLib-EAOD dataset covering diverse optical degradation patterns from different lens types, enabling models that generalize impressively across varied aberration distributions. We anticipate further work on universal blind aberration correction model designs building on this dataset.
The AODLib-EAOD data contains 3 parts. Here are the descriptions:
-Imgs: This training ready dataset comprises 2,111 pairs of degraded and ground truth images. Each pair is produced by… See the full description on the dataset page:
https://huggingface.co/datasets/Strange97/AODLib-EAOD.
