深度可分离卷积替换普通卷积
时间: 2023-10-16 21:06:43 浏览: 175
深度可分离卷积是一种将普通卷积过程分解为两个独立步骤的卷积方法。它由depthwise卷积和pointwise卷积组成。其中,depthwise卷积是在每个输入通道上独立进行的卷积操作,而pointwise卷积是在所有输入通道上进行的1x1卷积操作。通过这种方式,深度可分离卷积可以大大减少参数量和计算量,同时保持输出结果的一致性。因此,可以将深度可分离卷积用来替代普通卷积,以减少模型的复杂度和计算开销。<span class="em">1</span><span class="em">2</span><span class="em">3</span>
#### 引用[.reference_title]
- *1* [OSNet代码解析(普通卷积,分组卷积,深度可分离卷积,OSblock组成)](https://blog.csdn.net/m0_46204224/article/details/117747658)[target="_blank" data-report-click={"spm":"1018.2226.3001.9630","extra":{"utm_source":"vip_chatgpt_common_search_pc_result","utm_medium":"distribute.pc_search_result.none-task-cask-2~all~insert_cask~default-1-null.142^v93^chatsearchT3_1"}}] [.reference_item style="max-width: 50%"]
- *2* *3* [深度可分离卷积(DepthwiseSeparableConvolution):Depthwise卷积与Pointwise卷积](https://blog.csdn.net/qq_40243750/article/details/125958617)[target="_blank" data-report-click={"spm":"1018.2226.3001.9630","extra":{"utm_source":"vip_chatgpt_common_search_pc_result","utm_medium":"distribute.pc_search_result.none-task-cask-2~all~insert_cask~default-1-null.142^v93^chatsearchT3_1"}}] [.reference_item style="max-width: 50%"]
[ .reference_list ]
阅读全文