Table II.
# | Input | Operator | Output |
---|---|---|---|
1 | 2242 × 6 | 3 × 3 conv2d, stride 2 | 1122 × 48 |
2 | 1122 × 48 | 3 × 3 conv2d, stride 1 | 1122 × 48 |
3 | 1122 × 48 | 3 × 3 conv2d, stride 1 | 1122 × 48 |
4 | 1122 × 48 | 3 × 3 max pool, stride 2 | 562 × 48 |
b0 | 562 × 48 | Bottleneck block | 562 × 24 |
b1 | t(b0), 562 × 24 | Bottleneck block | 562 × 24 |
b2 | c[t(b0) – t(b1)], 562 × 48 | Bottleneck block | 562 × 24 |
b3 | c[t(b0) – t(b2)], 562 × 72 | Bottleneck block | 562 × 24 |
b4 | c[t(b0) – t(b3)], 282 × 96 | Bottleneck block | 282 × 24 |
b5 | c[t(b1) – t(b4)], 282 × 96 | Bottleneck block | 282 × 24 |
b6 | c[t(b2) – t(b5)], 282 × 96 | Bottleneck block | 282 × 24 |
b7 | c[t(b3) – t(b6)], 282 × 96 | Bottleneck block | 282 × 24 |
b8 | c[t(b4) – t(b7)], 142 × 96 | Bottleneck block | 142 × 24 |
b9 | c[t(b5) – t(b8)], 142 × 96 | Bottleneck block | 142 × 24 |
b10 | c[t(b6) – t(b9)], 142 × 96 | Bottleneck block | 142 × 24 |
b11 | c[t(b7) – t(b10)], 142 × 96 | Bottleneck block | 142 × 24 |
b12 | c[t(b8) – t(b11)], 72 × 96 | Bottleneck block | 72 × 24 |
b13 | c[t(b9) – t(b12)], 72 × 96 | Bottleneck block | 72 × 24 |
b14 | c[t(b10) – t(b13)], 72 × 96 | Bottleneck block | 72 × 24 |
b15 | c[t(b11) – t(b14)], 72 × 96 | Bottleneck block | 72 × 24 |
21 | c[b12 – b15], 72 × 96 | Global average pool | 96 |
22 | 96 | Fully connected layer | 2/3/4 |
c[t(b0) – t(b3)] means concatenate c[] each output of bottleneck blocks b0 to b3 after transfer block t().