We present the first dense, real-world, and large-scale dataset for intrinsic image decomposition. Our dataset was derived from the Multi-Illumination Dataset by Murmann et al. and consists of 1000 scenes under 25 different illuminations each, with 1000 unique albedo maps and 25.000 image - RGB shading pairs. This dataset accompanies our ACM Trans. Graph. 2023 paper.

We combine and extend existing flash/no-flash datasets to create a diverse set of real-world flash/ambient pairs suitable for training deep networks. Our dataset is constructed from three existing datasets: The Multi-Illumination Dataset (MID), The Flash and Ambient Illuminations Dataset (FAID), and the Deep Flash Portrait Dataset (DPD). We propose a pipeline for compositing portraits from DPD onto backgrounds from FAID, and discuss multiple considerations for normalizing and augmenting flash/no-flash data. This dataset accompanies our CVPR 2023 paper.

Web-based ineractive tool for editing monocular depth maps with real-time feedback on 3D geometry. Accompanies our SIGGRAPH 2022 Poster.

The Flash and Ambient Illuminations Dataset (FAID) consists of aligned flash-only and ambient-only illumination pairs captured with mobile devices by many crowd-workers participated in our collection effort. This dataset accompanies our ECCV 2018 paper.

This toolbox includes a collection of common affinity-based image matting algorithms as well as matte refinement methods used by sampling-based image matting methods. It features the only public (re-)implementation of information-flow matting, a faster implementation of matting Laplacian and a faster trimap trimming. The parameters for each algorithm are easily customizable. The toolbox is designed to be ease of use for an extended set of applications. Sparse affinity matrices defined and used in common matting method can be obtained by calling the corresponding functions. Please check the github page for more information and the source code.