SIGNet: Intrinsic Image Decomposition by a Semantic and Invariant Gradient Driven Network for Indoor Scenes

TL;DR

SIGNet introduces a hybrid deep learning method combining semantic and invariant priors with a progressive CNN to improve intrinsic image decomposition in indoor scenes, achieving state-of-the-art results.

Contribution

It proposes a novel combination of handcrafted priors and deep learning for more accurate IID, addressing dataset bias issues.

Findings

Achieves state-of-the-art performance on IID benchmarks.

Using priors and progressive CNN improves decomposition accuracy.

Code is publicly available for reproducibility.

Abstract

Intrinsic image decomposition (IID) is an under-constrained problem. Therefore, traditional approaches use hand crafted priors to constrain the problem. However, these constraints are limited when coping with complex scenes. Deep learning-based approaches learn these constraints implicitly through the data, but they often suffer from dataset biases (due to not being able to include all possible imaging conditions). In this paper, a combination of the two is proposed. Component specific priors like semantics and invariant features are exploited to obtain semantically and physically plausible reflectance transitions. These transitions are used to steer a progressive CNN with implicit homogeneity constraints to decompose reflectance and shading maps. An ablation study is conducted showing that the use of the proposed priors and progressive CNN increase the IID performance. State of the art performance on both our proposed dataset and the standard real-world IIW dataset shows the effectiveness of the proposed method. Code is made available at https://github.com/Morpheus3000/SIGNet