Multi-view multi-lighting captured dataset - Doll and Temple scenes - v1.0 JPEG, March 2014

This archive contains the JPEG version of two captured scenes ("Doll" and "Temple") introduced in the following paper:
	@Article{Laffont12,
	  author       = "Laffont, Pierre-Yves and Bousseau, Adrien and Paris, Sylvain and Durand, Fredo and Drettakis, George",
	  title        = "Coherent Intrinsic Images from Photo Collections",
	  journal      = "ACM Transactions on Graphics (Proc. SIGGRAPH Asia)",
	  volume       = "31",
	  year         = "2012",
	  url          = "http://www-sop.inria.fr/reves/Basilic/2012/LBPDD12"
	}
Please cite this paper if you use the dataset.

Both scenes were captured from multiple viewpoints and with multiple lighting conditions, in a lab environment. We used a Canon EOS 5D Mark II camera mounted on a tripod. 

For images named "lightingA_xxx" to "lightingG_xxx" we used two light sources: a remote-controlled flash which produces strong lighting, and a camera-mounted low-intensity flash which simulates ambient light in shadows (to prevent noisy dark regions). For images "lightingZ_xxx", only the camera-mounted low-intensity flash was used.
We captured multiple views of each scene (11 for the Doll scene and 10 for the Temple scene).

In addition, we captured an extra set of images ("xxx_viewpoint111" for Temple, and "xxx_viewpoint112" for Doll) after sticking pieces of gray cards onto the objects, in an attempt to measure the received illumination. Note that the objects were moved/deformed during the sticking, and as a result we did not use those images in our work; however, we share these extra images in the hope that they can be useful to the community.

All photographs were saved in Canon raw .CR2 format. They were converted using the dcraw tool ("dcraw -T -h -o 1") then saved to .JPG, resulting in 8-bit gamma-corrected JPEG images with 2817x1877 resolution. Note that the scaling factor applied during conversion varies across images; if this is a problem, use the TIFF version of this dataset instead.
