MB-TaylorFormer：基于泰勒公式扩展的多分支高效Transformer用于图像去雾
MB-TaylorFormer: Multi-branch Efficient Transformer Expanded by Taylor Formula for Image Dehazing
AI总结：
针对Transformer在图像去雾中计算复杂度过高的问题，提出MB-TaylorFormer，用泰勒展开近似softmax注意力实现线性复杂度，结合多尺度补丁嵌入的多分支架构，在多个基准上以轻量计算达到SOTA性能。
摘要：
近年来，Transformer网络因其全局感受野和对输入的适应性，开始在计算机视觉领域取代纯卷积神经网络（CNN）。然而，softmax注意力机制的二次方计算复杂度限制了其在图像去雾任务中的广泛应用，尤其是对于高分辨率图像。为了解决这一问题，我们提出了一种新的Transformer变体，该变体应用泰勒展开来近似softmax注意力，实现了线性计算复杂度。我们提出了一个多尺度注意力细化模块作为补充，以纠正泰勒展开的误差。此外，我们在所提出的Transformer中引入了一种具有多尺度补丁嵌入的多分支架构，该架构通过不同尺度的重叠可变形卷积来嵌入特征。多尺度补丁嵌入的设计基于三个关键思想：1）不同大小的感受野；2）多级语义信息；3）灵活形状的感受野。我们的模型，名为基于泰勒公式扩展的多分支Transformer（MB-TaylorFormer），能够在补丁嵌入阶段更灵活地嵌入从粗到细的特征，并以有限的计算成本捕获长距离像素交互。在多个去雾基准上的实验结果表明，MB-TaylorFormer以轻量计算负担实现了最先进的（SOTA）性能。源代码和预训练模型可在https://github.com/FVL2020/ICCV-2023-MB-TaylorFormer获取。
英文摘要：
In recent years, Transformer networks are beginning to replace pure convolutional neural networks (CNNs) in the field of computer vision due to their global receptive field and adaptability to input. However, the quadratic computational complexity of softmax-attention limits the wide application in image dehazing task, especially for high-resolution images. To address this issue, we propose a new Transformer variant, which applies the Taylor expansion to approximate the softmax-attention and achieves linear computational complexity. A multi-scale attention refinement module is proposed as a complement to correct the error of the Taylor expansion. Furthermore, we introduce a multi-branch architecture with multi-scale patch embedding to the proposed Transformer, which embeds features by overlapping deformable convolution of different scales. The design of multi-scale patch embedding is based on three key ideas: 1) various sizes of the receptive field; 2) multi-level semantic information; 3) flexible shapes of the receptive field. Our model, named Multi-branch Transformer expanded by Taylor formula (MB-TaylorFormer), can embed coarse to fine features more flexibly at the patch embedding stage and capture long-distance pixel interactions with limited computational cost. Experimental results on several dehazing benchmarks show that MB-TaylorFormer achieves state-of-the-art (SOTA) performance with a light computational burden. The source code and pre-trained models are available at https://github.com/FVL2020/ICCV-2023-MB-TaylorFormer.
机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。
