Magritte Sphere Video

 # Magritte-Sphere Video sequence by LISA ULB

The test sequence "Magritte Sphere Video" is provided by Sarah Fachada, Daniele Bonatto, Mehrdad Teratani, Gauthier Lafruit, members of the LISA department, EPB (Ecole Polytechnique de Bruxelles), ULB (Universite Libre de Bruxelles), Belgium.

 # License:

CC BY-NC-SA

 # Terms of Use:

Anykind of publication or report using this sequence should refer to the following references.

[1] Sarah Fachada, Daniele Bonatto, Mehrdad Teratani, Gauthier Lafruit, "Magritte Sphere Video Test Sequence", 2021.

@misc{fachada_magrittevideo_2021,
    title = {{Magritte} {Sphere} {Video} {Test} {Sequence}},
    author = {Fachada, Sarah and Bonatto, Daniele and Teratani, Mehrdad and Lafruit, Gauthier},
    month = feb,
    year = {2021},
    doi = {10.5281/zenodo.5048270}
}

[2] Sarah Fachada, Daniele Bonatto, Mehrdad Teratani, and Gauthier Lafruit, "Light Field Rendering for non-Lambertian Objects," presented at the Electronic Imaging, 2021.

@inproceedings{fachada_light_2021,
    title = {Light {Field} {Rendering} for non-{Lambertian} {Objects}},
    booktitle = {Electronic {Imaging}},
    author = {Fachada, Sarah and Bonatto, Daniele and Teratani, Mehrdad and Lafruit, Gauthier},
    year = {2021}
}

 # Production:

Laboratory of Image Synthesis and Analysis, LISA department, EPB, Universite Libre de Bruxelles, Belgium.

 # Content:

This dataset contains a test scene created and rendered with Blender [1] and the addon script [2] extended for Blender 2.8. We provide the Blender file and the rendered scene.

The scene contains a non-Lambertian (transparent-refractive (T) or mirror-specular (M)) sphere rendered in a regular camera array of 21x21 cameras. It describes a spiral around a central initial position within 17 frames.

In addition to the 3D model, we provide the rendered images : resolution of 2000x2000, the cameras are parallel, with a principal point at the center of the image.
We provide 17 frames of:
 - a regular subarray of 5x5 cameras (cameras number 66, 70, 74, 78, 82, 50, 154, 158, 162, 166, 234, 238, 242, 246, 250, 318, 322, 326, 330, 334, 402, 406, 410, 414, 418)
 - the central horizontal line of 21 cameras (cameras 210 to 230)

The dataset contains:
 - a `camera.json` file in OMAF coordinates system (Camera position: X: forwards, Y:left, Z: up, Rotation: yaw, pitch, roll) [3],
 - a `parameters.cfg` generated with [2],
 - a `texture_M` folder containing the rendered views in yuv420p10le format for the mirror object,
 - a `texture_T` folder containing the rendered views in yuv420p10le format for the transparent object,
 - a `mask` folder containing the mask indicating the sphere in yuv420p format,
 - a `depth_gt` folder containing the associated ground truth depth maps yuv420p16le format,
 - a `depth_estimated_M` folder containing the associated estimated depth maps yuv420p16le format,
 - a `depth_estimated_T` folder containing the associated estimated depth maps yuv420p16le format.
 
 
 # References and links:
 
[1] Blender Online Community, "Blender - a 3D modelling and rendering package." Blender Institute, Amsterdam: Blender Foundation, 2020.

[2] K. Honauer, O. Johannsen, D. Kondermann, and B. Goldluecke, "A Dataset and Evaluation Methodology for Depth Estimation on 4D Light Fields" in Asian Conference on Computer Vision, 2016,
https://github.com/lightfield-analysis/blender-addon
https://github.com/dbonattoj/blender-addon

[3] B. Kroon, "Reference View Synthesizer (RVS) manual [N18068]," ISO/IEC JTC1/SC29/WG11, Macau SAR, China, p. 19, Oct. 2018.
https://mpeg.chiariglione.org/standards/mpeg-i/omnidirectional-media-format