基于隐核的超卷积在医学成像中的应用
Hyper-Convolutions via Implicit Kernels for Medical Imaging
摘要：
卷积神经网络（CNN）是计算机视觉任务中最常用的架构之一。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的容量以及其性能与可学习的内核权重的数量直接相关，这是由通道数量和内核大小（支持）决定的。在本文中，我们提出了\textit{hyper-convolution}，一种新的构建块，隐式编码卷积核使用空间坐标。超卷积将内核大小与可学习参数的总数解耦，从而实现更灵活的架构设计。我们在实验中证明，用超卷积代替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：\emph{https://github.com/tym002/Hyper-Convolution}
英文摘要：
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 the \textit{hyper-convolution}, a novel building block that implicitly encodes the convolutional kernel using spatial coordinates. 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: \emph{https://github.com/tym002/Hyper-Convolution}
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。
