具有小波散射的通用深度网络

Dec, 2013

Generic Deep Networks with Wavelet Scattering

Edouard Oyallon, Stéphane Mallat, Laurent Sifre

TL;DR我们引入了一种两层小波散射网络，可用于物体分类，该两层卷积网络不涉及学习和最大池化，通过初始化第一层的小波滤波器，能够在形态物体变量和杂波等复杂图像数据集上高效执行。

Abstract

We introduce a two-layer wavelet scattering network, which involves no learning, for object classification. This scattering transform computes a spatial →