Kaggle hosts a dataset titled PGD_PNG_ATTACK_PER_SEGMENT_NOISE_GAN. The title suggests it contains images generated for adversarial attacks, likely using Projected Gradient Descent (PGD) and Generative Adversarial Networks (GANs). Metadata such as author, size, and license is unknown.
Use Cases
- Benchmarking model robustness against PGD-style adversarial attacks (inferred from domain, verify after download)
- Training or evaluating GAN-based adversarial example detectors (inferred from domain, verify after download)
- Studying the effects of per-segment noise perturbations on image classification (inferred from domain, verify after download)
Strengths
- Published on Kaggle, a major platform for data science resources.
Limitations
- Metadata is minimal; actual content requires verification after download.
- Column-level documentation is absent; field semantics must be inferred after download.
- Row count is unknown, which may limit suitability assessment.