Efficient Convolutional Network Learning using Parametric Log based Dual-Tree Wavelet ScatterNet

TL;DR

This paper introduces a hybrid CNN architecture that integrates a parametric log based DTCWT ScatterNet at the front, enhancing training efficiency and achieving competitive performance on standard image classification datasets.

Contribution

It presents a novel combination of ScatterNet with CNNs, replacing initial layers to improve learning efficiency and performance without pre-training.

Findings

Achieves comparable performance to pre-trained CNNs on CIFAR-10 and Caltech-101.

Demonstrates improved training efficiency due to edge-based invariant representations.

Shows competitive results against state-of-the-art methods.

Abstract

We propose a DTCWT ScatterNet Convolutional Neural Network (DTSCNN) formed by replacing the first few layers of a CNN network with a parametric log based DTCWT ScatterNet. The ScatterNet extracts edge based invariant representations that are used by the later layers of the CNN to learn high-level features. This improves the training of the network as the later layers can learn more complex patterns from the start of learning because the edge representations are already present. The efficient learning of the DTSCNN network is demonstrated on CIFAR-10 and Caltech-101 datasets. The generic nature of the ScatterNet front-end is shown by an equivalent performance to pre-trained CNN front-ends. A comparison with the state-of-the-art on CIFAR-10 and Caltech-101 datasets is also presented.