MarMaster/corruption-glass_blur

Corruption Dataset: Glass_Blur Dataset Description This dataset contains corrupted versions of ImageNet-1K images using glass_blur corruption. It is part of the ImageNet-C benchmark for evaluating model robustness to common image corruptions. Dataset Structure Train: 1,281,167 corrupted images Validation: 50,000 corrupted images Classes: 1000 ImageNet-1K classes Format: Arrow (Hugging Face Datasets) Corruption Type: Glass_Blur… See the full description on the dataset page: https://huggingface.co/datasets/MarMaster/corruption-glass_blur.