Self-Supervised Intrinsic Image Decomposition
This addresses the problem of data scarcity and ambiguity in intrinsic decomposition for computer vision applications, representing an incremental improvement over supervised methods.
The paper tackles the challenging problem of intrinsic image decomposition from a single image by proposing a self-supervised model that uses unsupervised reconstruction error to improve representations, enabling effective knowledge transfer to unseen categories, lighting, and shapes.
Intrinsic decomposition from a single image is a highly challenging task, due to its inherent ambiguity and the scarcity of training data. In contrast to traditional fully supervised learning approaches, in this paper we propose learning intrinsic image decomposition by explaining the input image. Our model, the Rendered Intrinsics Network (RIN), joins together an image decomposition pipeline, which predicts reflectance, shape, and lighting conditions given a single image, with a recombination function, a learned shading model used to recompose the original input based off of intrinsic image predictions. Our network can then use unsupervised reconstruction error as an additional signal to improve its intermediate representations. This allows large-scale unlabeled data to be useful during training, and also enables transferring learned knowledge to images of unseen object categories, lighting conditions, and shapes. Extensive experiments demonstrate that our method performs well on both intrinsic image decomposition and knowledge transfer.