Image Classification on ImageNet-1k (val) (Accuracy Delta and Compression Ratios)
-89Delta Top-1 Acc (%)NISP-50-B
Evaluation results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| NISP-50-BBackbone=ResNet-502025.12 | -89 | — | — | — | 1.78 | 1.79 | |
| DCPBackbone=ResNet-502025.12 | -106 | 76.01 | 92.93 | -56 | 2.06 | 2.25 | |
| NEXPBackbone=ResNet-50, compression_regime=Low2025.12 | -135 | 76.13 | 92.86 | -93 | 2 | 2.02 | |
| ThiNetBackbone=ResNet-502025.12 | -187 | 72.88 | 91.14 | -112 | 2.06 | 2.25 | |
| ABCBackbone=ResNet-502025.12 | -249 | 76.01 | 92.96 | -145 | 2.27 | 2.3 | |
| NEXPBackbone=ResNet-50, compression_regime=High2025.12 | -677 | 76.13 | 92.86 | -343 | 4.05 | 3.04 | |
| GAL-1-jointBackbone=ResNet-502025.12 | -684 | 76.15 | 92.87 | -375 | 2.5 | 3.68 | |
| HrankBackbone=ResNet-502025.12 | -715 | 76.15 | 92.87 | -329 | 3.08 | 4.17 |