Professor Paolo Favaro and PhD student Mehdi Noroozi are invited to the Rank Prize Symposium on Computer Vision, which is to be held in Grasmere, England, from August 20-23. It will bring together 10 senior and 20 early career researchers to discuss the topic of geometry and uncertainty in deep learning for computer vision. Among the speakers are Alyosha Efros, Raquel Urtasun, Vladlen Koltun, Daniel Cremers, Paolo Favaro, Andrew Blake, Andrea Vedaldi, Andrew Zisserman and Alex Kendall.

Prof. Paolo Favaro was invited to give a talk at the International Computer Vision Summer School (ICVSS) that took place in Sicily, Italy. The topic of this year’s school was “Computer Vision after Deep Learning”. In his talk “Beyond Supervised Learning”, Prof. Favaro presented state-of-the-art methods to extract features from data without labels, and methods to disentangle factors of variation in the data.

Recently, PhD student Simon Jenni visited the AI Summer School P.A.I.S.S. in Grenoble and received the award for best poster for his CVPR work Self-Supervised Feature Learning by Learning to Spot Artifacts. Congratulations!

We are very happy to announce that we have a total of eight accepted papers in the conferences CVPR, ECCV and SIGGRAPH. For CVPR we have a new record of four accepted submissions:

• Disentangling Factors of Variation by Mixing Them by Hu et al.
• Learning to Extract a Video Sequence from a Single Motion-Blurred Image by Jin et al.
• Self-Supervised Feature Learning by Learning to Spot Artifacts by Jenni et al.
• Boosting Self-Supervised Learning via Knowledge Transfer by Noroozi et al.

A common theme of all of these works is “self-supervised learning”, a trending topic in Computer Vision and Deep Learning that our group is very much interested in. Furthermore, in ECCV we have

• "Deep Bilevel Learning" by Jenni et al.,
• "Understanding Degeneracies and Ambiguities in Attribute Transfer" by Szabó et al.,
• and "Normalized Blind Deconvolution" by Jin et al.

We also have a SIGGRAPH work FaceShop: Deep Sketch-based Image Editing by Portenier et al. that got attention in the press. You can read the article here.