2021-3-11论文学习:ACNet及其若干魔改模块论文

ACNet: Strengthening the Kernel Skeletons for Powerful CNN via Asymmetric Convolution Blocks

论文地址：https://arxiv.org/pdf/1908.03930.pdf

论文收录于2019 IEEE/CVF International Conference on Computer Vision (ICCV)

代码地址：https://github.com/ShawnDing1994/ACNet

As designing appropriate Convolutional Neural Network(CNN) architecture in the context of a given application usually involves heavy human works or numerous GPU hours, the research community is soliciting the architecture-neutral CNN structures, which can be easily plugged into multiple mature architectures to improve the performance on our real-world applications. We propose Asymmetric Convolution Block (ACB), an architecture-neutral structure as a CNN building block, which uses 1D asymmetric convolutions to strengthen the square convolution kernels.For an off-the-shelf architecture, we replace the standard square-kernel convolutional layers with ACBs to construct an Asymmetric Convolutional Network (ACNet), which can be trained to reach a higher level of accuracy. After training, we equivalently convert the ACNet into the same original architecture,thus requiring no extra computations anymore. We have observed that ACNet can improve the performance of various models on CIFAR and ImageNet by a clear margin. Through further experiments, we attribute the effectiveness of ACB to its capability of enhancing the model’s robustness to rotational distortions and strengthening the central skeleton parts of square convolution kernels.

contributions

• 我们建议使用非对称卷积来显式地增强标准方核层的表示能力，非对称卷积可以融合到方核不需要额外的推理时间计算，而不是近似方核层像许多之前的工作[4,17,18,23,25,33]。
• 我们提出了ACB作为一种新的建筑中立的CNN构建块。我们可以通过简单地用ACB替换成熟架构中的每个平方核卷积层来构建一个ACN集，而不引入任何超参数，这样其有效性就可以与CNN架构设计文献中的许多进步相结合。
• 我们在CIFAR-10、CIFAR-100和图像网上的几个常见基准模型提高了明显的精度。
• 我们已经证明了骨架在标准方方形卷积核中的重要性，并证明了ACNE在增强这些骨架方面的有效性
• 我们已经证明了ACN特可以增强模型对旋转变形的鲁棒性，这可能会激发对旋转不变性问题的进一步研究。

在model中加入Horizontal和vertical卷积核之后，对于旋转后的样本，能增强对旋转不变性的分类

推荐知乎：https://zhuanlan.zhihu.com/p/131282789

评价

是一个即插即用的小模块，对于浅层网络的提升效果比较好，对于深层网络提升效果不好

Automatic Modulation Recognition for Radar Signals via Multi-Branch ACSE Networks

论文地址：https://sci-hub.se/10.1109/ACCESS.2020.2995203

论文收录于 IEEE ACCESS 2020

这不是我研究方向的论文，但是它引用了ACNet，我就想看看这里面到底是怎么在做文章。

总结来说文章提出了the multi-branch ACSE networks，网络结构如下：


不得不说这就是抄袭了，设计的网络就是把别人的ACnet和SE block合在了一起引入到了雷达信号领域，唯一的不同就是多分支，multi-branch。

只能说借鉴意义不大了。

• X. Ding, Y. Guo, G. Ding, and J. Han, ‘‘ACNet: Strengthening the kernel
  skeletons for powerful CNN via asymmetric convolution blocks,’’ in Proc.
  IEEE/CVF Int. Conf. Comput. Vis. (ICCV), Oct. 2019, pp. 1911–1920.

• Jie Hu, Li Shen, and Gang Sun. Squeeze-and-excitation networks. In Proceedings of the IEEE conference on computer vision and pattern recognition, pages 7132–7141, 2018.

ACSE Networks and Autocorrelation Features for PRI Modulation Recognition

论文地址：

IEEE COMMUNICATIONS LETTERS, VOL. 24, NO. 8, AUGUST 2020

提出了ACSE Networks，和上一篇文章一拨人搞得，真不知道有啥创新点

MMDM: Multi-frame and Multi-scale for Image Demoir´eing

论文地址：

2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW)

提出了一个 enhanced AC Block用于图像去噪。

相对于AC block多了两个个斜着的1*3卷积。

M2-Net: A Multi-scale Multi-level Feature Enhanced Network for Object Detection in Optical Remote Sensing Images

IEEE xplore上面居然只能看网页版没有pdf，就先不看了