MobileNet-v3详解

论文:《Searching for MobileNetV3》
论文链接:https://arxiv.org/abs/1905.02244v2
前文链接:SENet详解MobileNet-v1详解MobileNet-v2详解

■ 概述

       MobileNetV3发表于2019年,该V3版本结合了V1的深度可分离卷积、V2的Inverted Residuals和Linear Bottleneck以及SE Block,利用NAS(神经结构搜索)来搜索网络的配置和参数(关于V1、V2以及SE Block可以参考我的其他文章,链接已放到文章开头)。MobileNetV3提供了两个版本:MobileNetV3-Small 与 MobileNetV3-Large ,分别对应对计算和存储要求低和高的版本。同时,提出了一种新的高效分割解码器Lite reduce Atrous Spatial Pyramid Pooling (LR-ASPP)

h-swish激活函数

       Swish 激活函数已经被证明是一种比 ReLU 更佳的激活函数,但是相比 ReLU,它的计 算更复杂,因为有 sigmoid 函数。为了能够在移动设备上应用 swish 并降低它的计算开销, 提出了 h-swish。

● swish 激活函数:

在这里插入图片描述
● h-swish 激活函数:

在这里插入图片描述

● 图像

在这里插入图片描述

  • 0
    点赞
  • 1
    收藏
    觉得还不错? 一键收藏
  • 0
    评论
Mobilenet-V3 is a convolutional neural network architecture that is designed for efficient mobile and embedded vision applications. It is an extension of the Mobilenet-V2 architecture, with improvements in terms of accuracy, efficiency, and flexibility. To use Mobilenet-V3 in TensorFlow, you can either build the network from scratch or utilize pre-trained models provided by TensorFlow. The pre-trained models are available in different sizes and variations, such as small, large, and search-space options. To get started with Mobilenet-V3 in TensorFlow, you can follow these steps: 1. Install TensorFlow: Make sure you have TensorFlow installed on your system. You can use pip to install it: `pip install tensorflow`. 2. Load the Mobilenet-V3 model: If you want to use a pre-trained model, you can use the `tf.keras.applications` module to load it. For example, to load the small variant of Mobilenet-V3, you can use: ```python import tensorflow as tf model = tf.keras.applications.MobileNetV3Small() ``` 3. Preprocess your input data: Before feeding your data into the network, you may need to preprocess it according to the requirements of Mobilenet-V3. This typically involves resizing the images and normalizing the pixel values. 4. Run inference: Once you have loaded the model and preprocessed your data, you can run inference to obtain predictions. Pass your input data through the model and interpret the output based on your task (e.g., classification, object detection, etc.). Note that this is just a brief overview of how to use Mobilenet-V3 in TensorFlow. Depending on your specific requirements and use case, you may need to customize the architecture or fine-tune the pre-trained models. It's recommended to consult the TensorFlow documentation and resources for more detailed instructions and examples.
评论
添加红包

请填写红包祝福语或标题

红包个数最小为10个

红包金额最低5元

当前余额3.43前往充值 >
需支付:10.00
成就一亿技术人!
领取后你会自动成为博主和红包主的粉丝 规则
hope_wisdom
发出的红包
实付
使用余额支付
点击重新获取
扫码支付
钱包余额 0

抵扣说明:

1.余额是钱包充值的虚拟货币,按照1:1的比例进行支付金额的抵扣。
2.余额无法直接购买下载,可以购买VIP、付费专栏及课程。

余额充值