Research Article
An Electronic Component Recognition Algorithm Based on Deep Learning with a Faster SqueezeNet
Table 1
Model parameters for faster SqueezeNet.
| | Layer name/type | Output size | Filter size/stride | Depth | S1×1 (squeeze) | E1×1 (expand) | E3×3 (expand) |
| | Input | | — | — | — | — | — | | BatchNorm | | — | 0 | — | — | — | | Conv 1 | | | 1 | — | — | — | | Max pool 1 | | | 0 | — | — | — | | Fire 1 | | — | 2 | 8 | 16 | 32 | | Concat 1 | | — | 0 | — | — | — | | Conv 2 | | | 1 | — | — | — | | Max pool 2 | | | 0 | — | — | — | | Fire 2 | | — | 2 | 8 | 16 | 32 | | Concat 2 | | — | 0 | — | — | — | | Conv 3 | | | 1 | — | — | — | | Max pool 3 | | | 0 | — | — | — | | Fire 3 | | — | 2 | 8 | 16 | 32 | | Add | | — | 0 | — | — | — | | Conv 4 | | | 1 | — | — | — | | Conv 5 | | | 1 | — | — | — | | Global avgpool | | | 0 | — | — | — | | Softmax | 22 | — | — | — | — | — |
|
|
