Inception Transformer

Inception Transformer（iFormer）是一种新型的通用Transformer主干架构。它采用了通道划分机制，将卷积和最大池化与自注意力机制相结合，以便更好地捕捉高频信息，并扩展Transformer在频谱中的感知能力。此外，iFormer还引入了一个频率斜坡结构，利用灵活的Inception token混合器，在高频和低频分量之间进行有效的权衡。然而，该架构的一个明显限制是需要手动定义频率斜坡结构中的通道比率，这需要丰富的经验来更好地适应不同任务。<span class="em">1</span><span class="em">2</span><span class="em">3</span>
#### 引用[.reference_title]
- *1* [Inception Transformer：带有高低频混合注意力的视觉Vi](https://blog.csdn.net/m0_63642362/article/details/131485711)[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* [Inception Transformer解读](https://blog.csdn.net/qq_42499501/article/details/125272878)[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 ]