Intrinsic Images
in the Wild

Sean Bell, Kavita Bala, Noah Snavely

Cornell University

Intrinsic Images in the Wild is a large scale, public dataset for intrinsic image decompositions of real-world scenes selected from the OpenSurfaces dataset. Each image is annotated with crowdsourced pairwise comparisons of material properties. We develop a dense CRF-based algorithm for intrinsic image decomposition and show that it outperforms several state-of-the-art algorithms.

» publication, code and data, judgements, evaluation, decompositions.

Human Judgements

Reflectance

Shading

Human Judgements

Reflectance

Shading

Human Judgements

Reflectance

Shading