Image Classification on ImageNet (val) (Top-5 Accuracy and FLOPs Reduction)
89.08Top-5 AccuracyUncompressed
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| UncompressedModel=ResNet-18, Compression Method=None2021.07 | 89.08 | 1 | |
| ADMM-based TT-formatModel=ResNet-18, Compression Method=TT2021.07 | 89.08 | 2.47 | |
| FPGMModel=ResNet-18, Compression Method=Pruning2021.07 | 88.53 | 1.72 | |
| DSAModel=ResNet-18, Compression Method=Pruning2021.07 | 88.35 | 1.72 | |
| FBSModel=ResNet-18, Compression Method=Pruning2021.07 | 88.22 | 1.98 | |
| DACPModel=ResNet-18, Compression Method=Pruning2021.07 | 87.6 | 1.89 | |
| ADMM-based TT-formatModel=ResNet-18, Compression Method=TT2021.07 | 87.47 | 4.62 | |
| TRPModel=ResNet-18, Compression Method=Matrix2021.07 | 86.74 | 2.6 | |
| TRP+NuModel=ResNet-18, Compression Method=SVD2021.07 | 86.61 | 3.18 | |
| Standard TRModel=ResNet-18, Compression Method=TR2021.07 | 86.29 | 4.28 | |
| Standard TTModel=ResNet-18, Compression Method=TT2021.07 | 85.64 | 4.62 |