Hyper-convolutions via implicit kernels for medical image analysis
Hyper-convolutions via implicit kernels for medical image analysis
复制标题
DOI：
10.1016/j.media.2023.102796
复制
发表时间：
2023-03
影响因子：
10.9
通讯作者：
Tianyu Ma;Alan Q. Wang;Adrian V. Dalca;M. Sabuncu
中科院分区：
文献类型：
--
作者：
Tianyu Ma;Alan Q. Wang;Adrian V. Dalca;M. Sabuncu
卷积神经网络（CNN）是计算机视觉任务最常用的架构之一。 CNN 的关键构建块是卷积核，它聚合来自像素邻域的信息并在所有像素之间共享权重。标准 CNN 的容量及其性能与可学习的内核权重的数量直接相关，而可学习的内核权重由通道数和内核大小（支持度）决定。在本文中，我们提出了超卷积，这是一种使用空间坐标隐式编码卷积核的新颖构建块。与权重是独立学习的常规卷积核不同，超卷积核权重通过将空间坐标映射到相应值的编码器进行关联。超卷积将内核大小与可学习参数总数解耦，从而实现更灵活的架构设计。我们在实验中证明，用超卷积替换常规卷积可以用更少的参数提高性能，并提高对噪声的鲁棒性。我们在这里提供代码：https://github.com/tym002/Hyper-Convolution。
The convolutional neural network (CNN) is one of the most commonly used architectures for computer vision tasks. The key building block of a CNN is the convolutional kernel that aggregates information from the pixel neighborhood and shares weights across all pixels. A standard CNN’s capacity, and thus its performance, is directly related to the number of learnable kernel weights, which is determined by the number of channels and the kernel size (support). In this paper, we present thehyper-convolution, a novel building block that implicitly encodes the convolutional kernel using spatial coordinates. Unlike a regular convolutional kernel, whose weights are independently learned, hyper-convolution kernel weights are correlated through an encoder that maps spatial coordinates to their corresponding values. Hyper-convolutions decouple kernel size from the total number of learnable parameters, enabling a more flexible architecture design. We demonstrate in our experiments that replacing regular convolutions with hyper-convolutions can improve performance with less parameters, and increase robustness against noise. We provide our code here: https://github.com/tym002/Hyper-Convolution.