Dan Calian

Dan Calian Photo headshot

Before joining UCL's VEIV programme I obtained a BSc (Hons) in Computer Science with A.I. from the University of Nottingham. In my final year I worked on improving the performance of a genetics-based machine learning system through memetic extensions.

The goal of my EngD project is to investigate methods of realistically integrating virtually rendered content within real scenes. Achieving convincing results require knowing information about the real scene: at the very minimum it requires knowing the lighting conditions of the scene. For example, if these are known then virtual objects can be rendered to exhibit reflections and shadows which reasonably match those of real objects present in the real scene.

The problems of estimating material properties and/or lighting conditions of a real scene given only limited information about the scene (such as a photograph or short video) are collectively known as inverse rendering problems. Inverse rendering problems are often ill-posed, as they can have multiple solutions which agree with the given observations equally well. The correct solutions can only be distinguished through the use of additional prior information about the expected properties of these correct solutions. A significant part of my EngD lies in applying machine learning techniques to model the relevant properties of lighting conditions and geometry in forms amenable for use in solving these inverse rendering problems.

More specifically, I have investigated using custom 3D printed objects as means of acquiring the lighting conditions of real scenes (initially tackling only scenes composed of voxels using 3D-cross shaped objects and ultimately for tackling general scenes using specially designed spherical-partitions "shading probes"). I have also worked on estimating high-frequency high dynamic range outdoor lighting conditions from single photographs given an object of known class as well as on proposing and validating new formulations for general inverse lighting problems.


Conference papers

Jacques  Calian, Dan  Andrei  Calian,  Cristina  Amati,  Rebecca  Kleinberger,  Anthony  Steed,  Jan Kautz, and Tim Weyrich. 3D-Printing of non-assembly, articulated models. ACM Transactions on Graphics (Proc. SIGGRAPH Asia),  31(6):130:1-130:8, 2012.

Dan Andrei Calian, Kenny Mitchell, Derek Nowrouzezahrai, Jan Kautz. The Shading Probe: Fast Appearance Acquisition for Mobile AR. Proceedings of SIGGRAPH Asia 2013 Technical Briefs (SA 2013)  ACM, New York, NY, USA, Article 20.


Dan Andrei Calian, Kenny Mitchell. Optical Illumination Mapping. United States US20140267412. Publication date 18 September 2014.

Primary Supervisor: Prof Jan Kautz

Industry Sponsor: Disney Research

Inverse lighting | Augmented Reality | Geometry Optimisation | 3D Printing