Sill-Net: Feature Augmentation with Separated Illumination Representation

TL;DR

Sill-Net is a neural network that separates illumination features from images and uses this separation to augment training data, improving object recognition under varying illumination conditions.

Contribution

The paper introduces Sill-Net, a novel architecture that separates illumination features for data augmentation, enhancing recognition accuracy under diverse lighting.

Findings

Outperforms state-of-the-art methods in object classification benchmarks.

Effective separation of illumination features improves recognition robustness.

Data augmentation in feature space enhances model generalization.

Abstract

For visual object recognition tasks, the illumination variations can cause distinct changes in object appearance and thus confuse the deep neural network based recognition models. Especially for some rare illumination conditions, collecting sufficient training samples could be time-consuming and expensive. To solve this problem, in this paper we propose a novel neural network architecture called Separating-Illumination Network (Sill-Net). Sill-Net learns to separate illumination features from images, and then during training we augment training samples with these separated illumination features in the feature space. Experimental results demonstrate that our approach outperforms current state-of-the-art methods in several object classification benchmarks.