【轻量级网络】MobileNetV1: Efficient Convolutional Neural Networks for Mobile Vision Applications

---

---

MobileNetV1核心操作是深度可分离卷积

1、MobileNet Architecture

1.1、Depthwise Separable Convolution

标准卷积可以分解为两层：depthwise conv和pointwise conv

1.2、Network Structure and Training

2、Experiments

2.1、Model Choices

We next show results comparing thinner models with width multiplier to shallower models using less layers. To make MobileNet shallower, the 5 layers of separable filters with feature size 14 × 14 × 512 are removed.

2.2、Model Shrinking Hyperparameters