Computer Vision

Talks:
Searching on Manifolds - Ahmet Iscen (https://www.linkedin.com/in/ahmetius/)
This presentation focuses on similarity search on manifolds. The first part of the presentation investigates diffusion, a mechanism that captures the image manifold in the feature space. Despite the success of deep learning on representing images for instance-based image retrieval, recent studies show that the learned representations still lie on manifolds in a high dimensional space. Experimentally, we observe a significant boost in performance of image retrieval with compact CNN descriptors on standard benchmarks, especially when the query object covers only a small part of the image. The second part of the presentation focuses on hard training example mining for unsupervised metric learning. Experimentally, we show that our learned models are on par or are outperforming prior models that are fully or partially supervised.

Tolga Birdal (https://www.linkedin.com/in/tbirdal/)
With the advances in autonomous driving, robotics and geospatial mapping, utilization of raw 3D data in the form of point clouds, in order to perform SLAM, scene understanding or 3D reconstruction, started to attract tremendous attention from vision scholars. However, due to the unstructured and sparse nature of a point cloud, it is not immediate to adapt the powerful and well established tools of 2D computer vision to 3D domain. In defense of 3D geometric deep neural networks, Tolga Birdal argues for a learned local feature rather than a handcrafted one. In this talk, he will elaborate on the details of a set of novel proposals to strengthen local feature learning for 3D data. This talk will briefly introduce point cloud processing and later on combine findings of four publications.

Language: English

Speakers:
Ahmet Iscen (https://www.linkedin.com/in/ahmetius/)
Ahmet Iscen received his B.Sc. degree from the State University of New York at Binghamton in 2011, and his M.Sc. degree from Bilkent University in 2014. He completed his Ph.D. degree at Université de Rennes I, while working as a Ph.D. researcher at Inria Rennes. He joined Czech Technical University in Prague as a postdoctoral researcher in December 2017 and will join Google as a researcher scientist in January 2019. His work has appeared in selective conferences and journals, such as CVPR, ECCV, IEEE Transactions on Image Processing and IEEE Transactions on Big Data. He received the Fondation Rennes 1 Best Thesis Award 2017 in the field of Mathematics, Sciences and Information and Communication Technologies.

Tolga Birdal (http://tbirdal.me/)
Tolga Birdal is a PhD candidate at the Computer Vision Group, Chair for Computer Aided Medical Procedures, Technical University of Munich and a Doktorand at Siemens AG. He completed his Bachelors as an Electronics Engineer at the Sabanci University in 2008. In his subsequent postgraduate programme, he studied Computational Science and Engineering at Technical University of Munich. In continuation to his Master's thesis on “3D Deformable Surface Recovery Using RGBD Cameras”, he now focuses his research and development on large object detection, pose estimation and reconstruction using point clouds. Recently, he is awarded both Ernst von Siemens Scholarship and EMVA Young Professional Award 2016 for his PhD work. He has several publications at the well respected venues such as NIPS, CVPR, ECCV, ICCV, IROS, ICASSP and 3DV. Aside from his academic life, Tolga is a natural Entrepreneur. He has co-founded multiple companies including Befunky, a widely used web based image processing platform. For further information, visit tbirdal.me and http://campar.in.tum.de/Main/TolgaBirdal.

Program:

• 17:45 - 18:00 - Your arrival
• 18:00 - 18:40 - First talk
• 18:40 - 18:50 - Short break
• 18:50 - 19:30 - Second talk
• 19:30 - 22:00 - Networking in Bitcoin Coffee