Image Classification on ImageNet (val) (Top-1 Accuracy, Model Size)
81.6Top-1 AccUncompressed
Evaluation results
| Method | Links | ||
|---|---|---|---|
| UncompressedBackbone=EfficientNet-b32021.04 | 81.6 | 46.7 | |
| DiffQBackbone=EfficientNet-b3, Lambda (λ)=0.012021.04 | 81.5 | 8.7 | |
| QATBackbone=EfficientNet-b3, Bit-width=8 bits2021.04 | 81.3 | 12 | |
| DiffQBackbone=EfficientNet-b3, Lambda (λ)=0.052021.04 | 80.8 | 6 | |
| PQBackbone=EfficientNet-b32021.04 | 80 | 3.1 | |
| QATBackbone=EfficientNet-b3, Bit-width=4 bits2021.04 | 57.3 | 6.3 |