Learning rotation invariant convolutional filters for texture classification

TL;DR

This paper introduces a shallow CNN approach that learns rotation-invariant filters for texture classification, achieving high accuracy with fewer parameters by encoding rotation invariance directly into the model.

Contribution

The method encodes rotation invariance by tying filter weights to rotated versions, improving performance and reducing parameters compared to standard CNNs.

Findings

Achieves state-of-the-art or comparable results on texture classification benchmark.

Reduces the number of learned parameters by an order of magnitude.

Improves classification performance over standard shallow CNNs.

Abstract

We present a method for learning discriminative filters using a shallow Convolutional Neural Network (CNN). We encode rotation invariance directly in the model by tying the weights of groups of filters to several rotated versions of the canonical filter in the group. These filters can be used to extract rotation invariant features well-suited for image classification. We test this learning procedure on a texture classification benchmark, where the orientations of the training images differ from those of the test images. We obtain results comparable to the state-of-the-art. Compared to standard shallow CNNs, the proposed method obtains higher classification performance while reducing by an order of magnitude the number of parameters to be learned.