Name common objects from a clear image — what a typical adult recognizes instantly.
Recognize 20,000+ fine-grained categories under occlusion, rotation, and adversarial noise.
ImageNet classification at superhuman accuracy for over five years.
Adversarial robustness; out-of-distribution generalization to novel object composites.
| Technology | Quality | Source | Score vs. baseline | Score |
|---|---|---|---|---|
| No evidence records yet. | ||||