MNasNet
[1] Mingxing Tan, et al. MnasNet: Platform-Aware Neural Architecture Search for Mobile. CVPR 2019. Arxiv link: https://arxiv.org/pdf/1807.11626.pdf
About the Model
We provide a few standard-size and small-size AutoML models in mnasnet_models.py including:
- mnasnet-a1 has ~75.2% top-1 ImageNet accuracy with 3.9M parameters and 312M Multiply-Adds.
- mnasnet-small has ~66% top-1 ImageNet accuracy with 2.0M parameters and 68M Multiply-Adds.
The standard size MnasNet-A1 inference has 1.8x faster throughput (55% lower latency) than the corresponding MobileNetV2 model.
Comparing to MobileNetV2, MnasNet-A1 model has clear better performance in accuracy when they are at the same latency level.
Here are the details of Mnasnet-A1 on ImageNet:
| Input Size | Depth Multiplier | Top-1 Accuracy | Top-5 Accuracy | Parameters(M) | Multi-Adds (M) | Pixel 1 latency (ms) | |
|---|---|---|---|---|---|---|---|
| 224 | 1.4 | 77.2 | 93.5 | 6.1 | 591.5 | 135 | 77.2 |
| 224 | 1 | 75.2 | 92.5 | 3.9 | 315.2 | 78 | 75.2 |
| 224 | 0.75 | 73.3 | 91.3 | 2.9 | 226.7 | 61 | 73.3 |
| 224 | 0.5 | 68.9 | 88.4 | 2.1 | 105.2 | 32 | 68.9 |
| 224 | 0.35 | 64.1 | 85.1 | 1.7 | 63.2 | 22 | 64.1 |
| 192 | 1.4 | 76.1 | 93.0 | 6.1 | 435.1 | 99 | 76.1 |
| 192 | 1 | 74.0 | 91.6 | 3.9 | 232.0 | 57 | 74 |
| 192 | 0.75 | 72.1 | 90.5 | 2.9 | 166.9 | 45 | 72.1 |
| 192 | 0.5 | 67.2 | 87.4 | 2.1 | 77.6 | 24 | 67.2 |
| 192 | 0.35 | 62.4 | 83.8 | 1.7 | 46.8 | 17 | 62.4 |
| 160 | 1.4 | 74.8 | 92.1 | 6.1 | 302.8 | 72 | 74.8 |
| 160 | 1 | 72.0 | 90.5 | 3.9 | 161.6 | 41 | 72 |
| 160 | 0.75 | 70.1 | 89.3 | 2.9 | 116.4 | 33 | 70.1 |
| 160 | 0.5 | 64.9 | 85.8 | 2.1 | 54.4 | 18 | 64.9 |
| 160 | 0.35 | 52.3 | 81.5 | 1.7 | 32.9 | 13 | 59.3 |
| 128 | 1.4 | 72.5 | 90.6 | 6.1 | 194.5 | 49 | 72.5 |
| 128 | 1 | 69.3 | 88.9 | 3.9 | 104.1 | 29 | 69.3 |
| 128 | 0.75 | 67.0 | 87.3 | 2.9 | 75.0 | 23 | 67 |
| 128 | 0.5 | 60.8 | 83.0 | 2.1 | 35.3 | 12 | 60.8 |
| 128 | 0.35 | 54.8 | 78.1 | 1.7 | 21.6 | 8.5 | 54.8 |
| 96 | 1.4 | 68.6 | 88.1 | 6.1 | 110.3 | 32 | 68.6 |
| 96 | 1 | 64.4 | 85.8 | 3.9 | 59.3 | 18 | 64.4 |
| 96 | 0.75 | 62.1 | 84.0 | 2.9 | 42.9 | 17 | 62.1 |
| 96 | 0.5 | 54.7 | 78.1 | 2.1 | 20.5 | 7.4 | 54.7 |
| 96 | 0.35 | 49.3 | 73.4 | 1.7 | 12.7 | 5.4 | 49.3 |
For more information about training, please refer to our tutorial: https://cloud.google.com/tpu/docs/tutorials/mnasnet