Image Classification on ImageNet 640 samples (val)
196,900Kappa^2Standard Points
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| Standard PointsTile Size=F(8,3), Backbone=ResNet-18, Precision=FP16, Point Selection Method=Standard2025.12 | 196,900 | 4.7 | — | |
| Standard PointsTile Size=F(6,3), Backbone=ResNet-18, Precision=FP16, Point Selection Method=Standard2025.12 | 2,075 | 11.3 | — | |
| NOVATile Size=F(8,3), Backbone=ResNet-18, Precision=FP16, Point Selection Method=Discovered2025.12 | 474 | 50 | 45 | |
| NOVATile Size=F(6,3), Backbone=ResNet-18, Precision=FP16, Point Selection Method=Discovered2025.12 | 77 | 76.9 | 66 | |
| Standard PointsTile Size=F(4,3), Backbone=ResNet-18, Precision=FP16, Point Selection Method=Standard2025.12 | 42.5 | 78.3 | — | |
| NOVATile Size=F(4,3), Backbone=ResNet-18, Precision=FP16, Point Selection Method=Discovered2025.12 | 14.5 | 77.8 | 0 |