Image Classification on ImageNet (val) (Top-1 Error and Hardware Throughput)
24.6Top-1 ErrorShuffleNet v2 2x (with SE)
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| ShuffleNet v2 2x (with SE)Complexity (MFLOPs)=597, Resolution=224x224, Squeeze-and-Excitation=with SE [8]2018.07 | 24.6 | 161 | — | |
| ShuffleNet v2 2×Complexity (MFLOPs)=591, Resolution=224x2242018.07 | 25.1 | 217 | — | |
| MobileNet v2 1.4Complexity (MFLOPs)=585, Resolution=224x224, Scale=1.4, Implementation=original2018.07 | 25.3 | 137 | — | |
| IGCV3-D (1.4)Complexity (MFLOPs)=610, Resolution=224x224, Scale=1.42018.07 | 25.5 | 82 | — | |
| PNASNet-5Complexity (MFLOPs)=588, Resolution=224x224, Implementation=our implementation2018.07 | 25.8 | 115 | — | |
| NASNet-AComplexity (MFLOPs)=564, Resolution=224x224, Config=4 @ 1056, Implementation=our implementation2018.07 | 26 | 130 | — | |
| CondenseNet (G=C=4)Complexity (MFLOPs)=529, Resolution=224x224, Groups=G=C=42018.07 | 26.2 | — | — | |
| ShuffleNet v1 2×Complexity (MFLOPs)=524, Resolution=224x224, Groups=g=32018.07 | 26.3 | 133 | — | |
| MobileNet v2 1.4Complexity (MFLOPs)=587, Resolution=224x224, Scale=1.4, Implementation=our implementation2018.07 | 26.7 | 137 | — | |
| ShuffleNet v2 1.5×Complexity (MFLOPs)=299, Resolution=224x2242018.07 | 27.4 | 255 | — | |
| Xception 2×Complexity (MFLOPs)=525, Resolution=224x224, Implementation=our implementation2018.07 | 27.6 | 174 | — | |
| IGCV3-DComplexity (MFLOPs)=318, Resolution=224x2242018.07 | 27.8 | 102 | — | |
| MobileNet v2 1.0Complexity (MFLOPs)=300, Resolution=224x224, Scale=1.0, Implementation=original2018.07 | 28 | 180 | — | |
| MobileNet v2 1.0Complexity (MFLOPs)=301, Resolution=224x224, Scale=1.0, Implementation=our implementation2018.07 | 28.3 | 180 | — | |
| ShuffleNet v1 1.5×Complexity (MFLOPs)=292, Resolution=224x224, Groups=g=32018.07 | 28.5 | 164 | — | |
| CondenseNet (G=C=8)Complexity (MFLOPs)=274, Resolution=224x224, Groups=G=C=82018.07 | 29 | — | — | |
| IGCV2-1.0Complexity (MFLOPs)=564, Resolution=224x224, Scale=1.02018.07 | 29.3 | 81 | — | |
| Xception 1.5×Complexity (MFLOPs)=305, Resolution=224x224, Implementation=our implementation2018.07 | 29.4 | 219 | — | |
| MobileNet v1 1.0Complexity (MFLOPs)=569, Resolution=224x224, Scale=1.02018.07 | 29.4 | 247 | — | |
| ShuffleNet v2 1×Complexity (MFLOPs)=146, Resolution=224x2242018.07 | 30.6 | 341 | — | |
| IGCV3-D (0.7)Complexity (MFLOPs)=210, Resolution=224x224, Scale=0.72018.07 | 31.5 | 143 | — | |
| MobileNet v1 0.75Complexity (MFLOPs)=325, Resolution=224x224, Scale=0.75×2018.07 | 31.6 | 314 | — | |
| MobileNet v2 0.75Complexity (MFLOPs)=145, Resolution=192x192, Scale=0.75×, Implementation=our implementation2018.07 | 32.1 | 235 | — | |
| ShuffleNet v1 1×Complexity (MFLOPs)=140, Resolution=224x224, Groups=g=32018.07 | 32.6 | 213 | — | |
| MobileNet v2 0.6Complexity (MFLOPs)=141, Resolution=224x224, Scale=0.6×, Implementation=our implementation2018.07 | 33.3 | 249 | — | |
| Xception 1×Complexity (MFLOPs)=145, Resolution=224x224, Implementation=our implementation2018.07 | 34.1 | 278 | — | |
| IGCV2-0.5Complexity (MFLOPs)=156, Resolution=224x224, Scale=0.52018.07 | 34.5 | 132 | — | |
| DenseNet 2×Complexity (MFLOPs)=519, Resolution=224x224, Implementation=our implementation2018.07 | 34.6 | 197 | — | |
| MobileNet v1 0.5Complexity (MFLOPs)=149, Resolution=224x224, Scale=0.5×2018.07 | 36.3 | 382 | — | |
| ShuffleNet v2 0.5×Complexity (MFLOPs)=41, Resolution=224x224, Batch size (GPU/ARM)=8/12018.07 | 39.7 | 417 | — | |
| DenseNet 1.5×Complexity (MFLOPs)=295, Resolution=224x224, Implementation=our implementation2018.07 | 39.9 | 274 | — | |
| ShuffleNet v1 0.5×Complexity (MFLOPs)=38, Resolution=224x224, Groups=g=32018.07 | 43.2 | 347 | — | |
| MobileNet v2 0.4Complexity (MFLOPs)=43, Resolution=160x160, Scale=0.4×, Implementation=our implementation2018.07 | 43.4 | 333 | — | |
| Xception 0.5×Complexity (MFLOPs)=40, Resolution=224x224, Implementation=our implementation2018.07 | 44.9 | 384 | — | |
| IGCV2-0.25Complexity (MFLOPs)=46, Resolution=224x224, Scale=0.252018.07 | 45.1 | 183 | — | |
| DenseNet 1×Complexity (MFLOPs)=142, Resolution=224x224, Implementation=our implementation2018.07 | 45.2 | 279 | — | |
| MobileNet v1 0.25Complexity (MFLOPs)=41, Resolution=224x224, Scale=0.25×2018.07 | 49.4 | 502 | — | |
| MobileNet v2 0.15Complexity (MFLOPs)=39, Resolution=224x224, Scale=0.15×, Implementation=our implementation2018.07 | 55.1 | 351 | — | |
| DenseNet 0.5×Complexity (MFLOPs)=42, Resolution=224x224, Implementation=our implementation2018.07 | 58.6 | 366 | — |