Table 1. ShuffleNet architecture.
Layer | Output Size | Kernel Size | Stride | Repeat |
---|---|---|---|---|
image | 224×224 | |||
conv1 | 112×112 | 3×3 | 2 | 1 |
maxPool | 56×56 | 3×3 | 2 | |
stage2 | 28×28 | 2 | 1 | |
stage2 | 28×28 | 1 | 3 | |
stage3 | 14×14 | 2 | 1 | |
stage3 | 14×14 | 1 | 7 | |
stage4 | 7×7 | 2 | 1 | |
stage4 | 7×7 | 1 | 3 | |
globalPool | 1×1 | 7×7 | ||
fullyConnected | 2 | |||
softmax | 2 | |||
classification | 2 |