Inference Speed on Generic Synthetic Data
525GPU Throughput (B/s)ShuffleNet v2
Evaluation results
| Method | Links | ||
|---|---|---|---|
| ShuffleNet v2Input size=320x320, FLOPs=140M, Batch size=12018.07 | 525 | 12.5 | |
| ShuffleNet v2Input size=320x320, FLOPs=300M, Batch size=12018.07 | 474 | 6.1 | |
| XceptionInput size=320x320, FLOPs=140M, Batch size=12018.07 | 463 | 10.1 | |
| MobileNet v2Input size=320x320, FLOPs=140M, Batch size=12018.07 | 460 | 6.4 | |
| ShuffleNet v2Input size=640x480, FLOPs=40M, Batch size=12018.07 | 424 | 9.3 | |
| ShuffleNet v2Input size=320x320, FLOPs=500M, Batch size=12018.07 | 422 | 3.4 | |
| ShuffleNet v1Input size=320x320, FLOPs=140M, Batch size=12018.07 | 414 | 11.4 | |
| XceptionInput size=320x320, FLOPs=300M, Batch size=12018.07 | 408 | 5.6 | |
| XceptionInput size=640x480, FLOPs=40M, Batch size=12018.07 | 399 | 9.6 | |
| ShuffleNet v1Input size=640x480, FLOPs=40M, Batch size=12018.07 | 396 | 8 | |
| ShuffleNet v2Input size=640x480, FLOPs=140M, Batch size=12018.07 | 394 | 4 | |
| MobileNet v2Input size=320x320, FLOPs=300M, Batch size=12018.07 | 389 | 4.6 | |
| XceptionInput size=320x320, FLOPs=500M, Batch size=12018.07 | 350 | 3.5 | |
| ShuffleNet v1Input size=320x320, FLOPs=300M, Batch size=12018.07 | 344 | 5.1 | |
| MobileNet v2Input size=640x480, FLOPs=40M, Batch size=12018.07 | 338 | 3.8 | |
| MobileNet v2Input size=320x320, FLOPs=500M, Batch size=12018.07 | 335 | 2.7 | |
| XceptionInput size=640x480, FLOPs=140M, Batch size=12018.07 | 326 | 3.2 | |
| ShuffleNet v2Input size=320x320, FLOPs=40M, Batch size=82018.07 | 315 | 28.1 | |
| ShuffleNet v2Input size=640x480, FLOPs=300M, Batch size=12018.07 | 297 | 1.9 | |
| XceptionInput size=320x320, FLOPs=40M, Batch size=82018.07 | 279 | 31.1 | |
| ShuffleNet v1Input size=320x320, FLOPs=500M, Batch size=12018.07 | 275 | 3.1 | |
| ShuffleNet v1Input size=640x480, FLOPs=140M, Batch size=12018.07 | 269 | 3.7 | |
| ShuffleNet v2Input size=640x480, FLOPs=500M, Batch size=12018.07 | 250 | 1.1 | |
| MobileNet v2Input size=640x480, FLOPs=140M, Batch size=12018.07 | 248 | 2 | |
| ShuffleNet v2Input size=1080x720, FLOPs=40M, Batch size=12018.07 | 248 | 3.5 | |
| XceptionInput size=640x480, FLOPs=300M, Batch size=12018.07 | 244 | 1.7 | |
| ShuffleNet v1Input size=320x320, FLOPs=40M, Batch size=82018.07 | 236 | 27.2 | |
| XceptionInput size=1080x720, FLOPs=40M, Batch size=12018.07 | 232 | 3.6 | |
| XceptionInput size=640x480, FLOPs=500M, Batch size=12018.07 | 209 | 1.1 | |
| MobileNet v2Input size=640x480, FLOPs=300M, Batch size=12018.07 | 208 | 1.4 | |
| ShuffleNet v1Input size=1080x720, FLOPs=40M, Batch size=12018.07 | 203 | 2.9 | |
| ShuffleNet v1Input size=640x480, FLOPs=300M, Batch size=12018.07 | 198 | 1.6 | |
| ShuffleNet v2Input size=1080x720, FLOPs=140M, Batch size=12018.07 | 197 | 1.5 | |
| MobileNet v2Input size=320x320, FLOPs=40M, Batch size=82018.07 | 187 | 11.4 | |
| MobileNet v2Input size=640x480, FLOPs=500M, Batch size=12018.07 | 165 | 0.8 | |
| XceptionInput size=1080x720, FLOPs=140M, Batch size=12018.07 | 160 | 1.2 | |
| MobileNet v2Input size=1080x720, FLOPs=40M, Batch size=12018.07 | 159 | 1.4 | |
| ShuffleNet v1Input size=640x480, FLOPs=500M, Batch size=12018.07 | 156 | 1 | |
| ShuffleNet v2Input size=1080x720, FLOPs=300M, Batch size=12018.07 | 141 | 0.7 | |
| ShuffleNet v1Input size=1080x720, FLOPs=140M, Batch size=12018.07 | 131 | 1.4 | |
| XceptionInput size=1080x720, FLOPs=300M, Batch size=12018.07 | 124 | 0.5 | |
| MobileNet v2Input size=1080x720, FLOPs=140M, Batch size=12018.07 | 117 | 0.7 | |
| ShuffleNet v2Input size=1080x720, FLOPs=500M, Batch size=12018.07 | 115 | 0.4 | |
| XceptionInput size=1080x720, FLOPs=500M, Batch size=12018.07 | 106 | 0.4 | |
| MobileNet v2Input size=1080x720, FLOPs=300M, Batch size=12018.07 | 99 | 0.3 | |
| ShuffleNet v1Input size=1080x720, FLOPs=300M, Batch size=12018.07 | 96 | 0.4 | |
| MobileNet v2Input size=1080x720, FLOPs=500M, Batch size=12018.07 | 78 | 0.3 | |
| ShuffleNet v1Input size=1080x720, FLOPs=500M, Batch size=12018.07 | 77 | 0.3 |