Silvano Galliani

Ars est celare artem

News

May 2019: I received the outstanding reviewer award at CVPR 2019!
Jul 2018: I am currently a Scientist in the brand new Microsoft Hololens lab in Zurich!
Jun 2018: I won with Charis Lanaras the 2nd Place award in NTIRE 2018 Challenge on Spectral Reconstruction from RGB Images at CVPR 2018.
May 2018: I started as a PostDoc in the computer vision group at ETH with Prof. Marc Pollefeys .
May 2018: I received the outstanding reviewer award at CVPR 2018!
Jan 2018: Here it is! A new CNN network for sparse data: faster than tensorflow and reduced footprint. The paper is on Arxiv with full implementation in tensorflow. ILA-SCNN: Inference, Learning and Attention Mechanisms that Exploit and Preserve Sparsity in Convolutional Networks.
Dec 2017: I successfully defended my PhD Thesis with title "Multi-View 3D reconstruction with geometry and shading"!
Sep 2017: I'm honored te be invited at the prestigious Deep Learning in Computer Vision seminar in Dagstuhl, Germany.
Jul 2017: Learned Multi-patch Similarity is accepted at ICCV 2017, it's dedicated to the memory of Wilfried Hartmann.
Jul 2017: Our multi-view stereo / 3D reconstruction benchmark website is online now!
Mar 2017: Learned Spectral Super-Resolution is on arXiv: Spectral super-resolution by means of a semantic segmentation network: from RGB we predict 31 channnels. State of the art and cool results also on satellite data plus denoise as a byproduct.
Mar 2017: Learned Multi-patch Similarity is on arXiv: A multiple branch cnn siamese network learns multi-patch similarity across a varying number of patches.
Feb 2017: A Multi-View Stereo Benchmark with High-Resolution Images and Multi-Camera Videos accepted at CVPR17.
We released on arxiv our work on semanting segmentation that surpasses 90% accuracy on the ISPRS Vahingen benchmark
May 2016: I'll make a summer internship in Virtual Reality at Oculus research (officially for Facebook) in Pittsburgh.
10 Nov 2015: I gave a talk at the Swiss Machine Learning Day 2015
Paper accepted to CVPR 2016 as spotlight oral:
Silvano Galliani, Konrad Schindler: Just look at the image: viewpoint-specific surface normal prediction for improved multi-view reconstruction, CVPR 2016
My super fast and accurate multi-view stereo reconstruction algorithm has been released as free software on github, feel free to (ab)use it!
Currently I am a PhD student at ETH Zurich in the group of Konrad Schindler. Yay! ~~Check my new research web page for some up to date informations.~~

Publications

On my Google Scholar profile or:

T. Hackel, M. Usvyatsov, S. Galliani, J.D. Wegner, K. Schindler:
ILA-SCNN: Inference, Learning and Attention Mechanisms that Exploit and Preserve Sparsity in Convolutional Networks
Arxiv preprint, 2018. Code
D. Marmanis, K. Schindler, J.D. Wegner, S. Galliani, M. Datcu, U. Stilla:
Classification With an Edge: Improving Semantic Image Segmentation with Boundary Detection
ISPRS Journal of Photogrammetry and Remote Sensing, 2018
S. Galliani, C. Lanaras, D. Marmanis, E. Baltsavias, K. Schindler:
Learned Spectral Super-Resolution
Arxiv preprint 2017 (under review)
W. Hartmann, S. Galliani, M. Havlena, L.V. Gool, K. Schindler:
Learned Multi-patch Similarity
ICCV 2017
T. Schöps, J. Schönberger, S. Galliani, T. Sattler, K. Schindler, A. Geiger and M. Pollefeys
A Multi-View Stereo Benchmark with High-Resolution Images and Multi-Camera Videos in Unstructured Scenes
CVPR 2017
S. Galliani, K. Schindler:
Just look at the image: viewpoint-specific surface normal prediction for improved multi-view reconstruction
Spotlight oral CVPR 2016
D. Marmanis, J. Wegner, S. Galliani, S. Konrad, M. Datcu, U. Stilla:
Semantic Segmentation of Aerial Images with an Ensemble of Fully Convolutional Neural Networks
Oral ISPRS Congress 2016
S. Galliani, K. Lasinger and K. Schindler:
Massively Parallel Multiview Stereopsis by Surface Normal Diffusion, supplementary material
ICCV 2015 Code
S. Galliani, Y. C. Ju, M. Breuß, A. Bruhn:
Oral Generalised perspective shape from shading in spherical coordinates.
Scale Space and Variational Methods in Computer Vision (SSVM) 2013
S. Galliani, M. Breuß, Y.C. Ju:
Fast and robust surface normal integration by a discrete eikonal equation.
BMVC 2012.
Y.C. Ju, M. Breuß, A. Bruhn, S. Galliani:
Shape from shading for rough surfaces: Analysis of the Oren-Nayar model. BMVC 2012.
S. Galliani, Perspective Photometric Stereo with Non-Lambertian Reflectance Function, Master's Thesis in Visual Computing, Saarland University, Saarbrücken, Germany, December 2011.

Born in Milan, B.Sc. in Computer Science in Milan (Universita' di Milano Bicocca) while working (tvblob, mala and videomag), traveling around ,hacking and programming. Free software brought me in Norway, Brasil, Spain.

M.Sc. in Visual Computing in Saarbruecken (DE). I started and dropped a PhD student on Shape from Shading/Photometric stereo with the Fraunhofer institute in Kaiserslautern and BTU Cottbus.

I finished my PhD student at ETH Zurich, Switzerland, in the group of Kondrad Schindler. Check my research web page.

Art & Free software

Projects I did in the past. Recently I've been quite busy with research to continue them.

Freej (2003-2008)

I'm a developer of FreeJ. It is a vision mixer: an instrument for real-time video manipulation used in the fields of dance theater, veejaying, medical visualisation and TV. Most of all I contributed with the text layer and ogg/theora video streaming using icecast. (c++ language).

The Ball in The Hole (2006-2008)

The ball in the hole is a social installation/performance where the user becomes the actor of metamorphosis. The discovering of the self and the legendary mirror to other dimensions is experienced through a real-time tactile digital environment. The Alter is someone else, and the deformation shown is a constant force that the user can apply to his virtual figure.

Hands On (2005)

"Hands on" it's a funny installation I did back in 2005 in Stecca degli artigiani , Milan. Together with Matteo Erenbourg and Giacomo Porfiri, we collected hundreds of photo of hands and projected them in an evolving video during the same night. It was very funny. Recently I released all of the 600 photos under a Creative Common licence. Feel free do download them and use them according to the terms of the licence. I did also a video showing the result (the images were projected upside down).

Free software (2003-now)

I did collaborate to some free software project. It happened with the bittorrent client transmission, the icecast audio video streaming server and recently also n2n, an interesting p2p virtual private network bypassing firewall using popular skype technique of hole punching.

In the Past: Talks and Workshops

In the past I did many presentations and workshop on video streaming, freej and free software all around Europe.. Take a look at them.

(06-2004) audio streaming in Prague and in Freistadt (Austria) during fm@dia
(11-2004) freej presentation in Bergen (Norway) during pik5el festival
(10-2004) vim advanced in Milan (Italy) in reload reality hacking
(06-2005) freej video streaming in Tarifa (Spain) during fadaiat festival
(05-2007) complete audio video production only with free software in Venice inside cittadellarte
video streaming in Landgoed Velder (Netherland) during whatthehack camp
(10-2005) video streaming in Bergen (Norway) during pik5el meeting
(2006) video production and distribution with free software (mostly cinelerra and freej) in Spoleto (Italy) inside offucina
(2007) The ball in the hole presentation in San Paolo (Brasil) during File Electronic Arts festival

Contacts and Digital Identities

Email: kysucix@dyne.org
Facebook: ~~I am not there.~~ I'm there after the internship at Facebook Reality Labs but without "friends"
Twitter: Follow me, my name is kysucix.
Linkedin: I'm there as well.
Identi.ca: I used to write tips or cryptic sentences there, dead now.
Google+: I'm not a google employee :D
Google: My alter-ego is kysucix, google me if you're curious.
Firefox: Browse the list of firefox addons I'm currently using.
Github: Of course I have a github account