<div dir="ltr"><div dir="ltr"><div class="gmail_default" style="font-size:small"><div class="gmail_default"><div class="gmail_default"><p style="font-variant-numeric:normal;font-variant-east-asian:normal;font-stretch:normal;line-height:normal;margin:0px"><font face="arial, sans-serif"><font style="vertical-align:inherit"><font style="vertical-align:inherit"><b>When:</b>    </font></font><font style="vertical-align:inherit"><font style="vertical-align:inherit">  Tuesday, January 26th at<b> 11:10 am CT</b></font></font><br></font></p><p class="MsoNormal" style="margin:0in 0in 0.0001pt;line-height:normal;background-image:initial;background-position:initial;background-size:initial;background-repeat:initial;background-origin:initial;background-clip:initial"><font face="arial, sans-serif"> </font></p><p class="MsoNormal" style="margin:0in 0in 0.0001pt;line-height:normal;background-image:initial;background-position:initial;background-size:initial;background-repeat:initial;background-origin:initial;background-clip:initial"><font face="arial, sans-serif"><font style="vertical-align:inherit"><font style="vertical-align:inherit"><b>Where:</b>     </font></font></font><font color="#000000" style="font-family:arial,sans-serif">Zoom Virtual Talk (</font><b style="font-family:arial,sans-serif"><font color="#0000ff"><a href="https://uchicagogroup.zoom.us/webinar/register/WN_KJ4Af7uDQqadp5OgFOS6OQ" target="_blank">register in advance here</a></font></b><font color="#000000" style="font-family:arial,sans-serif">)</font></p><p class="MsoNormal" style="margin:0in 0in 0.0001pt;line-height:normal;background-image:initial;background-position:initial;background-size:initial;background-repeat:initial;background-origin:initial;background-clip:initial"><font face="arial, sans-serif"> </font></p><p class="MsoNormal" style="margin:0in 0in 0.0001pt;line-height:normal;background-image:initial;background-position:initial;background-size:initial;background-repeat:initial;background-origin:initial;background-clip:initial"><font face="arial, sans-serif"><font style="vertical-align:inherit"><font style="vertical-align:inherit"><b>Who: </b>       </font></font></font>Tolga Birdal, Stanford University</p></div><br></div><div class="gmail_default"><div style="color:rgb(0,0,0)"><div style="margin:0px"><b><span style="margin:0px"><b><font face="arial, sans-serif"><span style="margin:0px">Title</span><span style="margin:0px"><span style="margin:0px;font-weight:400;color:rgb(102,102,102)">: Non-Euclidean Machine Learning for 3D Computer Vision</span></span></font></b><br></span></b></div><div style="font-size:12pt;font-family:Calibri,Arial,Helvetica,sans-serif;margin:0px"><b><span style="margin:0px"><b><span style="margin:0px"><span style="margin:0px;font-weight:400;font-size:14px;font-family:"Open Sans","Helvetica Neue",Helvetica,Arial,sans-serif;color:rgb(102,102,102)"><span style="margin:0px"><br></span></span></span></b></span></b></div><div style="margin:0px"><font face="arial, sans-serif"><b><span style="margin:0px"><span style="margin:0px">Abstract</span></span></b>: <span style="margin:0px;color:rgb(102,102,102)">We understand the world by interacting with the objects and agents we observe. This Kantian empirical realism called</span><span style="margin:0px;color:rgb(102,102,102)"><span style="margin:0px"> </span></span><span style="margin:0px;color:rgb(102,102,102)"><i style="box-sizing:border-box">experience</i></span><span style="margin:0px;color:rgb(102,102,102)"><span style="margin:0px"> </span></span><span style="margin:0px;color:rgb(102,102,102)">is made possible by the a priori Euclidean constraints on space. While being subject to limits of scales and tolerances of our senses, such a</span><span style="margin:0px;color:rgb(102,102,102)"><span style="margin:0px"> </span></span><span style="margin:0px;color:rgb(102,102,102)"><strong style="box-sizing:border-box">flat</strong></span><span style="margin:0px;color:rgb(102,102,102)"><span style="margin:0px"> </span></span><span style="margin:0px;color:rgb(102,102,102)">view of the world has been the driving force in many engineering fields laying the foundations of the first AI systems. For example, </span><span style="margin:0px;color:rgb(102,102,102)"><span style="margin:0px">the typical data that</span><span style="margin:0px"><span style="margin:0px"> </span><span style="margin:0px">learning</span><span style="margin:0px"> </span></span><span style="margin:0px">systems consume such as text, audio, or images are organized into a</span><span style="margin:0px"><span style="margin:0px"> </span></span><span style="margin:0px"><b>grid/lattice</b></span><span style="margin:0px"><span style="margin:0px"> </span></span><span style="margin:0px">just like the pixel matrix underlying an image. This makes the processing easy and allows researchers to devise domain specific algorithms. </span><span style="margin:0px"><span style="margin:0px">On the other hand, the typical output of a</span><span style="margin:0px"><span style="margin:0px"> </span><span style="margin:0px">learning</span><span style="margin:0px"> </span><span style="margin:0px">machine</span><span style="margin:0px"> </span></span><span style="margin:0px">is a set of real valued numbers that best explain a downstream task such as predicting the weather temperature. Neural networks as the de-facto choices are then responsible for optimally mapping the space of the input to the output space, both of which are</span><span style="margin:0px"><span style="margin:0px"> </span></span><span style="margin:0px"><b>assumed Euclidean</b></span><span style="margin:0px">. </span><span style="margin:0px"><span style="margin:0px">Unfortunately, for a variety of applications ranging from 3D computer vision to analysis of social networks or protein structures, the assumptions posed by Euclidean geometry cease to hold. F</span><span style="margin:0px"><span style="margin:0px">or instance,</span><span style="margin:0px"><span style="margin:0px"> </span></span><span style="margin:0px"><b>3D surfaces</b></span><span style="margin:0px">,</span><span style="margin:0px"><b><span style="margin:0px"> </span></b></span><span style="margin:0px"><b>point clouds</b></span><span style="margin:0px">,</span><span style="margin:0px"><span style="margin:0px"> </span></span><span style="margin:0px"><b>trees</b></span><span style="margin:0px"><span style="margin:0px"> </span></span><span style="margin:0px">or</span><span style="margin:0px"><span style="margin:0px"> </span></span><span style="margin:0px"><b>graphs</b></span><span style="margin:0px"><span style="margin:0px"> </span></span><span style="margin:0px">are types of inputs whose data points neither satisfy the Pythagorean theorem nor can be arranged into a grid without information loss. Furthermore, entities such as</span><span style="margin:0px"><span style="margin:0px"> </span></span><span style="margin:0px"><b>rotations</b></span><span style="margin:0px">,</span><span style="margin:0px"><span style="margin:0px"> </span></span><span style="margin:0px"><b>permutations</b></span><span style="margin:0px"><span style="margin:0px"> </span></span><span style="margin:0px">(orderings) or</span><span style="margin:0px"><span style="margin:0px"> </span></span><span style="margin:0px"><b>probability distributions</b></span><span style="margin:0px"><span style="margin:0px"> </span></span><span style="margin:0px">cannot be regressed (or predicted) without restricting the domain of real numbers. In other words, they lie on a lower dimensional</span><span style="margin:0px"><span style="margin:0px"> </span></span><span style="margin:0px"><b>sub-manifold</b></span><span style="margin:0px"><span style="margin:0px"> </span></span><span style="margin:0px">endowed with a certain geometric structure. </span></span></span></span></span></font></div><div style="margin:0px"><span style="margin:0px;color:rgb(102,102,102)"><span style="margin:0px"><span style="margin:0px"><span style="margin:0px"><span style="margin:0px"><span style="margin:0px"><font face="arial, sans-serif"><br></font></span></span></span></span></span></span></div><div style="margin:0px"><span style="margin:0px;color:rgb(102,102,102)"><span style="margin:0px"><span style="margin:0px"><span style="margin:0px"><span style="margin:0px"><font face="arial, sans-serif"><span style="margin:0px">In my research, I challenge this Euclidean perspective and propose to work on the </span><span style="margin:0px"><span style="margin:0px">non-Euclidean</span></span><span style="margin:0px">, curved structure of the environments that surround us.</span><span style="margin:0px"><span style="margin:0px"> </span><span style="margin:0px"><span style="margin:0px">I coin such mapping of</span><span style="margin:0px"><span style="margin:0px"> </span><span style="margin:0px">non-Euclidean</span><span style="margin:0px"> </span></span><span style="margin:0px">input to</span><span style="margin:0px"><span style="margin:0px"> </span><span style="margin:0px">non-Euclidean</span><span style="margin:0px"> </span></span><span style="margin:0px">output the</span><span style="margin:0px"><span style="margin:0px"> </span></span><span style="margin:0px"><b><span style="margin:0px">non-Euclidean</span><span style="margin:0px"> </span><span style="margin:0px">machine</span><span style="margin:0px"> </span><span style="margin:0px">learning</span></b></span><span style="margin:0px">. </span><span style="margin:0px"><span style="margin:0px">My ultimate goal is to arm the autonomous systems running on 3D data with capabilities of scene or object-level reasoning natively on the</span><span style="margin:0px"><span style="margin:0px"> </span></span><span style="margin:0px"><b>geometric</b></span><span style="margin:0px"><span style="margin:0px"> </span></span><span style="margin:0px">nature of the 3D perception problem. In this talk, I will summarize my previous research on processing 3D point clouds to enable understanding of rigid and non-rigid dynamics. I will also investigate how to provide the additional</span><span style="margin:0px"><span style="margin:0px"> </span></span><span style="margin:0px"><b>uncertainty</b></span><span style="margin:0px"><span style="margin:0px"> </span></span><span style="margin:0px">information for the problems at hand. Finally, I will open a window into the future technologies and applications such approach can enable. </span></span></span></span></font></span></span></span></span></span></div><font face="arial, sans-serif"><br></font></div><div style="color:rgb(0,0,0)"><font face="arial, sans-serif"><b><span style="margin:0px"><span style="margin:0px">BIO</span></span></b>: <span style="margin:0px;color:rgb(102,102,102)">Tolga Birdal is a Postdoctoral Research Fellow at Stanford University. He carries his research within the Geometric Computing Group of Prof. Leonidas Guibas. Previously, Tolga has defended his PhD thesis at the Computer Vision Group, Chair for Computer Aided Medical Procedures, Technical University of Munich led by Prof. Nassir Navab. He was also a Doktorand at Siemens AG. Tolga completed his Bachelors as an Electronics Engineer at 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 focused his research and development on large object detection, pose estimation and reconstruction using point clouds. Tolga 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 NeurIPS, CVPR, ICCV, ECCV, 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 <a href="http://tbirdal.me/" target="_blank">tbirdal.me</a>, </span><span style="margin:0px;color:rgb(102,102,102)"><a href="https://profiles.stanford.edu/tolga-birdal" target="_blank">https://profiles.stanford.edu/tolga-birdal</a></span><span style="margin:0px;color:rgb(102,102,102)">.</span></font></div><div id="gmail-m_-931738937922747375gmail-m_6606618964153194863gmail-m_6013610420488477655gmail-m_-6312913977678378189gmail-m_2026403581502646114appendonsend"></div><br></div><div class="gmail_default"><p style="color:rgb(60,64,67);letter-spacing:0.2px;white-space:pre-wrap"><b style="font-family:arial,sans-serif;letter-spacing:normal;color:rgb(34,34,34)">Host:</b><span style="font-family:arial,sans-serif;letter-spacing:normal;color:rgb(34,34,34)"> <a href="mailto:mwalter@ttic.edu" target="_blank">Matthew Walter</a></span><br></p><br class="gmail-Apple-interchange-newline"></div></div><div><div dir="ltr" class="gmail_signature" data-smartmail="gmail_signature"><div dir="ltr"><div><div dir="ltr"><div><div dir="ltr"><div><div dir="ltr"><div><div dir="ltr"><div><div dir="ltr"><div><div dir="ltr"><font face="arial, helvetica, sans-serif">Mary C. Marre</font><div><font face="arial, helvetica, sans-serif">Faculty Administrative Support</font></div><div><i><font face="arial, helvetica, sans-serif" color="#3d85c6"><b>Toyota Technological Institute</b></font></i></div><div><i><font face="arial, helvetica, sans-serif" color="#3d85c6">6045 S. Kenwood Avenue</font></i></div><div><i><font face="arial, helvetica, sans-serif" color="#3d85c6">Room 517</font></i></div><div><i><font face="arial, helvetica, sans-serif" color="#3d85c6">Chicago, IL  60637</font></i></div><div><i><font face="arial, helvetica, sans-serif">p:(773) 834-1757</font></i></div><div><i><font face="arial, helvetica, sans-serif">f: (773) 357-6970</font></i></div><div><b><i><a href="mailto:mmarre@ttic.edu" target="_blank"><font face="arial, helvetica, sans-serif">mmarre@ttic.edu</font></a></i></b></div></div></div></div></div></div></div></div></div></div></div></div></div></div></div></div><br></div><br><div class="gmail_quote"><div dir="ltr" class="gmail_attr">On Mon, Jan 25, 2021 at 4:02 PM Mary Marre <<a href="mailto:mmarre@ttic.edu">mmarre@ttic.edu</a>> wrote:<br></div><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex"><div dir="ltr"><div dir="ltr"><div style="font-size:small"><div><div><p style="font-variant-numeric:normal;font-variant-east-asian:normal;font-stretch:normal;line-height:normal;margin:0px"><font face="arial, sans-serif"><font style="vertical-align:inherit"><font style="vertical-align:inherit"><b>When:</b>    </font></font><font style="vertical-align:inherit"><font style="vertical-align:inherit">  Tuesday, January 26th at<b> 11:10 am CT</b></font></font><br></font></p><p class="MsoNormal" style="margin:0in 0in 0.0001pt;line-height:normal;background-image:initial;background-position:initial;background-size:initial;background-repeat:initial;background-origin:initial;background-clip:initial"><font face="arial, sans-serif"> </font></p><p class="MsoNormal" style="margin:0in 0in 0.0001pt;line-height:normal;background-image:initial;background-position:initial;background-size:initial;background-repeat:initial;background-origin:initial;background-clip:initial"><font face="arial, sans-serif"><font style="vertical-align:inherit"><font style="vertical-align:inherit"><b>Where:</b>     </font></font></font><font color="#000000" style="font-family:arial,sans-serif">Zoom Virtual Talk (</font><b style="font-family:arial,sans-serif"><font color="#0000ff"><a href="https://uchicagogroup.zoom.us/webinar/register/WN_KJ4Af7uDQqadp5OgFOS6OQ" target="_blank">register in advance here</a></font></b><font color="#000000" style="font-family:arial,sans-serif">)</font></p><p class="MsoNormal" style="margin:0in 0in 0.0001pt;line-height:normal;background-image:initial;background-position:initial;background-size:initial;background-repeat:initial;background-origin:initial;background-clip:initial"><font face="arial, sans-serif"> </font></p><p class="MsoNormal" style="margin:0in 0in 0.0001pt;line-height:normal;background-image:initial;background-position:initial;background-size:initial;background-repeat:initial;background-origin:initial;background-clip:initial"><font face="arial, sans-serif"><font style="vertical-align:inherit"><font style="vertical-align:inherit"><b>Who: </b>       </font></font></font>Tolga Birdal, Stanford University</p></div><br></div><div><div style="color:rgb(0,0,0)"><div style="margin:0px"><b><span style="margin:0px"><b><font face="arial, sans-serif"><span style="margin:0px">Title</span><span style="margin:0px"><span style="margin:0px;font-weight:400;color:rgb(102,102,102)">: Non-Euclidean Machine Learning for 3D Computer Vision</span></span></font></b><br></span></b></div><div style="font-size:12pt;font-family:Calibri,Arial,Helvetica,sans-serif;margin:0px"><b><span style="margin:0px"><b><span style="margin:0px"><span style="margin:0px;font-weight:400;font-size:14px;font-family:"Open Sans","Helvetica Neue",Helvetica,Arial,sans-serif;color:rgb(102,102,102)"><span style="margin:0px"><br></span></span></span></b></span></b></div><div style="margin:0px"><font face="arial, sans-serif"><b><span style="margin:0px"><span style="margin:0px">Abstract</span></span></b>: <span style="margin:0px;color:rgb(102,102,102)">We understand the world by interacting with the objects and agents we observe. This Kantian empirical realism called</span><span style="margin:0px;color:rgb(102,102,102)"><span style="margin:0px"> </span></span><span style="margin:0px;color:rgb(102,102,102)"><i style="box-sizing:border-box">experience</i></span><span style="margin:0px;color:rgb(102,102,102)"><span style="margin:0px"> </span></span><span style="margin:0px;color:rgb(102,102,102)">is made possible by the a priori Euclidean constraints on space. While being subject to limits of scales and tolerances of our senses, such a</span><span style="margin:0px;color:rgb(102,102,102)"><span style="margin:0px"> </span></span><span style="margin:0px;color:rgb(102,102,102)"><strong style="box-sizing:border-box">flat</strong></span><span style="margin:0px;color:rgb(102,102,102)"><span style="margin:0px"> </span></span><span style="margin:0px;color:rgb(102,102,102)">view of the world has been the driving force in many engineering fields laying the foundations of the first AI systems. For example, </span><span style="margin:0px;color:rgb(102,102,102)"><span style="margin:0px">the typical data that</span><span style="margin:0px"><span style="margin:0px"> </span><span style="margin:0px">learning</span><span style="margin:0px"> </span></span><span style="margin:0px">systems consume such as text, audio, or images are organized into a</span><span style="margin:0px"><span style="margin:0px"> </span></span><span style="margin:0px"><b>grid/lattice</b></span><span style="margin:0px"><span style="margin:0px"> </span></span><span style="margin:0px">just like the pixel matrix underlying an image. This makes the processing easy and allows researchers to devise domain specific algorithms. </span><span style="margin:0px"><span style="margin:0px">On the other hand, the typical output of a</span><span style="margin:0px"><span style="margin:0px"> </span><span style="margin:0px">learning</span><span style="margin:0px"> </span><span style="margin:0px">machine</span><span style="margin:0px"> </span></span><span style="margin:0px">is a set of real valued numbers that best explain a downstream task such as predicting the weather temperature. Neural networks as the de-facto choices are then responsible for optimally mapping the space of the input to the output space, both of which are</span><span style="margin:0px"><span style="margin:0px"> </span></span><span style="margin:0px"><b>assumed Euclidean</b></span><span style="margin:0px">. </span><span style="margin:0px"><span style="margin:0px">Unfortunately, for a variety of applications ranging from 3D computer vision to analysis of social networks or protein structures, the assumptions posed by Euclidean geometry cease to hold. F</span><span style="margin:0px"><span style="margin:0px">or instance,</span><span style="margin:0px"><span style="margin:0px"> </span></span><span style="margin:0px"><b>3D surfaces</b></span><span style="margin:0px">,</span><span style="margin:0px"><b><span style="margin:0px"> </span></b></span><span style="margin:0px"><b>point clouds</b></span><span style="margin:0px">,</span><span style="margin:0px"><span style="margin:0px"> </span></span><span style="margin:0px"><b>trees</b></span><span style="margin:0px"><span style="margin:0px"> </span></span><span style="margin:0px">or</span><span style="margin:0px"><span style="margin:0px"> </span></span><span style="margin:0px"><b>graphs</b></span><span style="margin:0px"><span style="margin:0px"> </span></span><span style="margin:0px">are types of inputs whose data points neither satisfy the Pythagorean theorem nor can be arranged into a grid without information loss. Furthermore, entities such as</span><span style="margin:0px"><span style="margin:0px"> </span></span><span style="margin:0px"><b>rotations</b></span><span style="margin:0px">,</span><span style="margin:0px"><span style="margin:0px"> </span></span><span style="margin:0px"><b>permutations</b></span><span style="margin:0px"><span style="margin:0px"> </span></span><span style="margin:0px">(orderings) or</span><span style="margin:0px"><span style="margin:0px"> </span></span><span style="margin:0px"><b>probability distributions</b></span><span style="margin:0px"><span style="margin:0px"> </span></span><span style="margin:0px">cannot be regressed (or predicted) without restricting the domain of real numbers. In other words, they lie on a lower dimensional</span><span style="margin:0px"><span style="margin:0px"> </span></span><span style="margin:0px"><b>sub-manifold</b></span><span style="margin:0px"><span style="margin:0px"> </span></span><span style="margin:0px">endowed with a certain geometric structure. </span></span></span></span></span></font></div><div style="margin:0px"><span style="margin:0px;color:rgb(102,102,102)"><span style="margin:0px"><span style="margin:0px"><span style="margin:0px"><span style="margin:0px"><span style="margin:0px"><font face="arial, sans-serif"><br></font></span></span></span></span></span></span></div><div style="margin:0px"><span style="margin:0px;color:rgb(102,102,102)"><span style="margin:0px"><span style="margin:0px"><span style="margin:0px"><span style="margin:0px"><font face="arial, sans-serif"><span style="margin:0px">In my research, I challenge this Euclidean perspective and propose to work on the </span><span style="margin:0px"><span style="margin:0px">non-Euclidean</span></span><span style="margin:0px">, curved structure of the environments that surround us.</span><span style="margin:0px"><span style="margin:0px"> </span><span style="margin:0px"><span style="margin:0px">I coin such mapping of</span><span style="margin:0px"><span style="margin:0px"> </span><span style="margin:0px">non-Euclidean</span><span style="margin:0px"> </span></span><span style="margin:0px">input to</span><span style="margin:0px"><span style="margin:0px"> </span><span style="margin:0px">non-Euclidean</span><span style="margin:0px"> </span></span><span style="margin:0px">output the</span><span style="margin:0px"><span style="margin:0px"> </span></span><span style="margin:0px"><b><span style="margin:0px">non-Euclidean</span><span style="margin:0px"> </span><span style="margin:0px">machine</span><span style="margin:0px"> </span><span style="margin:0px">learning</span></b></span><span style="margin:0px">. </span><span style="margin:0px"><span style="margin:0px">My ultimate goal is to arm the autonomous systems running on 3D data with capabilities of scene or object-level reasoning natively on the</span><span style="margin:0px"><span style="margin:0px"> </span></span><span style="margin:0px"><b>geometric</b></span><span style="margin:0px"><span style="margin:0px"> </span></span><span style="margin:0px">nature of the 3D perception problem. In this talk, I will summarize my previous research on processing 3D point clouds to enable understanding of rigid and non-rigid dynamics. I will also investigate how to provide the additional</span><span style="margin:0px"><span style="margin:0px"> </span></span><span style="margin:0px"><b>uncertainty</b></span><span style="margin:0px"><span style="margin:0px"> </span></span><span style="margin:0px">information for the problems at hand. Finally, I will open a window into the future technologies and applications such approach can enable. </span></span></span></span></font></span></span></span></span></span></div><font face="arial, sans-serif"><br></font></div><div style="color:rgb(0,0,0)"><font face="arial, sans-serif"><b><span style="margin:0px"><span style="margin:0px">BIO</span></span></b>: <span style="margin:0px;color:rgb(102,102,102)">Tolga Birdal is a Postdoctoral Research Fellow at Stanford University. He carries his research within the Geometric Computing Group of Prof. Leonidas Guibas. Previously, Tolga has defended his PhD thesis at the Computer Vision Group, Chair for Computer Aided Medical Procedures, Technical University of Munich led by Prof. Nassir Navab. He was also a Doktorand at Siemens AG. Tolga completed his Bachelors as an Electronics Engineer at 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 focused his research and development on large object detection, pose estimation and reconstruction using point clouds. Tolga 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 NeurIPS, CVPR, ICCV, ECCV, 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 <a href="http://tbirdal.me/" target="_blank">tbirdal.me</a>, </span><span style="margin:0px;color:rgb(102,102,102)"><a href="https://profiles.stanford.edu/tolga-birdal" target="_blank">https://profiles.stanford.edu/tolga-birdal</a></span><span style="margin:0px;color:rgb(102,102,102)">.</span></font></div><div id="gmail-m_-931738937922747375gmail-m_6606618964153194863gmail-m_6013610420488477655gmail-m_-6312913977678378189gmail-m_2026403581502646114appendonsend"></div><br></div><div><p style="color:rgb(60,64,67);letter-spacing:0.2px;white-space:pre-wrap"><b style="font-family:arial,sans-serif;letter-spacing:normal;color:rgb(34,34,34)">Host:</b><span style="font-family:arial,sans-serif;letter-spacing:normal;color:rgb(34,34,34)"> <a href="mailto:mwalter@ttic.edu" target="_blank">Matthew Walter</a></span><br></p><p style="font-size:14px;color:rgb(60,64,67);font-family:Roboto,Arial,sans-serif;letter-spacing:0.2px;white-space:pre-wrap"><br style="letter-spacing:0.2px">

</p></div></div><div><div dir="ltr"><div dir="ltr"><div><div dir="ltr"><div><div dir="ltr"><div><div dir="ltr"><div><div dir="ltr"><div><div dir="ltr"><div><div dir="ltr"><font face="arial, helvetica, sans-serif">Mary C. Marre</font><div><font face="arial, helvetica, sans-serif">Faculty Administrative Support</font></div><div><i><font face="arial, helvetica, sans-serif" color="#3d85c6"><b>Toyota Technological Institute</b></font></i></div><div><i><font face="arial, helvetica, sans-serif" color="#3d85c6">6045 S. Kenwood Avenue</font></i></div><div><i><font face="arial, helvetica, sans-serif" color="#3d85c6">Room 517</font></i></div><div><i><font face="arial, helvetica, sans-serif" color="#3d85c6">Chicago, IL  60637</font></i></div><div><i><font face="arial, helvetica, sans-serif">p:(773) 834-1757</font></i></div><div><i><font face="arial, helvetica, sans-serif">f: (773) 357-6970</font></i></div><div><b><i><a href="mailto:mmarre@ttic.edu" target="_blank"><font face="arial, helvetica, sans-serif">mmarre@ttic.edu</font></a></i></b></div></div></div></div></div></div></div></div></div></div></div></div></div></div></div></div><br></div><br><div class="gmail_quote"><div dir="ltr" class="gmail_attr">On Thu, Jan 21, 2021 at 2:55 PM Mary Marre <<a href="mailto:mmarre@ttic.edu" target="_blank">mmarre@ttic.edu</a>> wrote:<br></div><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex"><div dir="ltr"><div><div style="font-size:small"><div><p style="font-variant-numeric:normal;font-variant-east-asian:normal;font-stretch:normal;line-height:normal;margin:0px"><font face="arial, sans-serif"><font style="vertical-align:inherit"><font style="vertical-align:inherit"><b>When:</b>    </font></font><font style="vertical-align:inherit"><font style="vertical-align:inherit">  Tuesday, January 26th at<b> 11:10 am CT</b></font></font><br></font></p><p class="MsoNormal" style="margin:0in 0in 0.0001pt;line-height:normal;background-image:initial;background-position:initial;background-size:initial;background-repeat:initial;background-origin:initial;background-clip:initial"><font face="arial, sans-serif"> </font></p><p class="MsoNormal" style="margin:0in 0in 0.0001pt;line-height:normal;background-image:initial;background-position:initial;background-size:initial;background-repeat:initial;background-origin:initial;background-clip:initial"><font face="arial, sans-serif"><font style="vertical-align:inherit"><font style="vertical-align:inherit"><b>Where:</b>     </font></font></font><font color="#000000" style="font-family:arial,sans-serif">Zoom Virtual Talk (</font><b style="font-family:arial,sans-serif"><font color="#0000ff"><a href="https://uchicagogroup.zoom.us/webinar/register/WN_KJ4Af7uDQqadp5OgFOS6OQ" target="_blank">register in advance here</a></font></b><font color="#000000" style="font-family:arial,sans-serif">)</font></p><p class="MsoNormal" style="margin:0in 0in 0.0001pt;line-height:normal;background-image:initial;background-position:initial;background-size:initial;background-repeat:initial;background-origin:initial;background-clip:initial"><font face="arial, sans-serif"> </font></p><p class="MsoNormal" style="margin:0in 0in 0.0001pt;line-height:normal;background-image:initial;background-position:initial;background-size:initial;background-repeat:initial;background-origin:initial;background-clip:initial"><font face="arial, sans-serif"><font style="vertical-align:inherit"><font style="vertical-align:inherit"><b>Who: </b>       </font></font></font>Tolga Birdal, Stanford University</p></div><br></div><div><div style="color:rgb(0,0,0)"><div style="margin:0px"><b><span style="margin:0px"><b><font face="arial, sans-serif"><span style="margin:0px">Title</span><span style="margin:0px"><span style="margin:0px;font-weight:400;color:rgb(102,102,102)">: Non-Euclidean Machine Learning for 3D Computer Vision</span></span></font></b><br></span></b></div><div style="font-size:12pt;font-family:Calibri,Arial,Helvetica,sans-serif;margin:0px"><b><span style="margin:0px"><b><span style="margin:0px"><span style="margin:0px;font-weight:400;font-size:14px;font-family:"Open Sans","Helvetica Neue",Helvetica,Arial,sans-serif;color:rgb(102,102,102)"><span style="margin:0px"><br></span></span></span></b></span></b></div><div style="margin:0px"><font face="arial, sans-serif"><b><span style="margin:0px"><span style="margin:0px">Abstract</span></span></b>: <span style="margin:0px;color:rgb(102,102,102)">We understand the world by interacting with the objects and agents we observe. This Kantian empirical realism called</span><span style="margin:0px;color:rgb(102,102,102)"><span style="margin:0px"> </span></span><span style="margin:0px;color:rgb(102,102,102)"><i style="box-sizing:border-box">experience</i></span><span style="margin:0px;color:rgb(102,102,102)"><span style="margin:0px"> </span></span><span style="margin:0px;color:rgb(102,102,102)">is made possible by the a priori Euclidean constraints on space. While being subject to limits of scales and tolerances of our senses, such a</span><span style="margin:0px;color:rgb(102,102,102)"><span style="margin:0px"> </span></span><span style="margin:0px;color:rgb(102,102,102)"><strong style="box-sizing:border-box">flat</strong></span><span style="margin:0px;color:rgb(102,102,102)"><span style="margin:0px"> </span></span><span style="margin:0px;color:rgb(102,102,102)">view of the world has been the driving force in many engineering fields laying the foundations of the first AI systems. For example, </span><span style="margin:0px;color:rgb(102,102,102)"><span style="margin:0px">the typical data that</span><span style="margin:0px"><span style="margin:0px"> </span><span style="margin:0px">learning</span><span style="margin:0px"> </span></span><span style="margin:0px">systems consume such as text, audio, or images are organized into a</span><span style="margin:0px"><span style="margin:0px"> </span></span><span style="margin:0px"><b>grid/lattice</b></span><span style="margin:0px"><span style="margin:0px"> </span></span><span style="margin:0px">just like the pixel matrix underlying an image. This makes the processing easy and allows researchers to devise domain specific algorithms. </span><span style="margin:0px"><span style="margin:0px">On the other hand, the typical output of a</span><span style="margin:0px"><span style="margin:0px"> </span><span style="margin:0px">learning</span><span style="margin:0px"> </span><span style="margin:0px">machine</span><span style="margin:0px"> </span></span><span style="margin:0px">is a set of real valued numbers that best explain a downstream task such as predicting the weather temperature. Neural networks as the de-facto choices are then responsible for optimally mapping the space of the input to the output space, both of which are</span><span style="margin:0px"><span style="margin:0px"> </span></span><span style="margin:0px"><b>assumed Euclidean</b></span><span style="margin:0px">. </span><span style="margin:0px"><span style="margin:0px">Unfortunately, for a variety of applications ranging from 3D computer vision to analysis of social networks or protein structures, the assumptions posed by Euclidean geometry cease to hold. F</span><span style="margin:0px"><span style="margin:0px">or instance,</span><span style="margin:0px"><span style="margin:0px"> </span></span><span style="margin:0px"><b>3D surfaces</b></span><span style="margin:0px">,</span><span style="margin:0px"><b><span style="margin:0px"> </span></b></span><span style="margin:0px"><b>point clouds</b></span><span style="margin:0px">,</span><span style="margin:0px"><span style="margin:0px"> </span></span><span style="margin:0px"><b>trees</b></span><span style="margin:0px"><span style="margin:0px"> </span></span><span style="margin:0px">or</span><span style="margin:0px"><span style="margin:0px"> </span></span><span style="margin:0px"><b>graphs</b></span><span style="margin:0px"><span style="margin:0px"> </span></span><span style="margin:0px">are types of inputs whose data points neither satisfy the Pythagorean theorem nor can be arranged into a grid without information loss. Furthermore, entities such as</span><span style="margin:0px"><span style="margin:0px"> </span></span><span style="margin:0px"><b>rotations</b></span><span style="margin:0px">,</span><span style="margin:0px"><span style="margin:0px"> </span></span><span style="margin:0px"><b>permutations</b></span><span style="margin:0px"><span style="margin:0px"> </span></span><span style="margin:0px">(orderings) or</span><span style="margin:0px"><span style="margin:0px"> </span></span><span style="margin:0px"><b>probability distributions</b></span><span style="margin:0px"><span style="margin:0px"> </span></span><span style="margin:0px">cannot be regressed (or predicted) without restricting the domain of real numbers. In other words, they lie on a lower dimensional</span><span style="margin:0px"><span style="margin:0px"> </span></span><span style="margin:0px"><b>sub-manifold</b></span><span style="margin:0px"><span style="margin:0px"> </span></span><span style="margin:0px">endowed with a certain geometric structure. </span></span></span></span></span></font></div><div style="margin:0px"><span style="margin:0px;color:rgb(102,102,102)"><span style="margin:0px"><span style="margin:0px"><span style="margin:0px"><span style="margin:0px"><span style="margin:0px"><font face="arial, sans-serif"><br></font></span></span></span></span></span></span></div><div style="margin:0px"><span style="margin:0px;color:rgb(102,102,102)"><span style="margin:0px"><span style="margin:0px"><span style="margin:0px"><span style="margin:0px"><font face="arial, sans-serif"><span style="margin:0px">In my research, I challenge this Euclidean perspective and propose to work on the </span><span style="margin:0px"><span style="margin:0px">non-Euclidean</span></span><span style="margin:0px">, curved structure of the environments that surround us.</span><span style="margin:0px"><span style="margin:0px"> </span><span style="margin:0px"><span style="margin:0px">I coin such mapping of</span><span style="margin:0px"><span style="margin:0px"> </span><span style="margin:0px">non-Euclidean</span><span style="margin:0px"> </span></span><span style="margin:0px">input to</span><span style="margin:0px"><span style="margin:0px"> </span><span style="margin:0px">non-Euclidean</span><span style="margin:0px"> </span></span><span style="margin:0px">output the</span><span style="margin:0px"><span style="margin:0px"> </span></span><span style="margin:0px"><b><span style="margin:0px">non-Euclidean</span><span style="margin:0px"> </span><span style="margin:0px">machine</span><span style="margin:0px"> </span><span style="margin:0px">learning</span></b></span><span style="margin:0px">. </span><span style="margin:0px"><span style="margin:0px">My ultimate goal is to arm the autonomous systems running on 3D data with capabilities of scene or object-level reasoning natively on the</span><span style="margin:0px"><span style="margin:0px"> </span></span><span style="margin:0px"><b>geometric</b></span><span style="margin:0px"><span style="margin:0px"> </span></span><span style="margin:0px">nature of the 3D perception problem. In this talk, I will summarize my previous research on processing 3D point clouds to enable understanding of rigid and non-rigid dynamics. I will also investigate how to provide the additional</span><span style="margin:0px"><span style="margin:0px"> </span></span><span style="margin:0px"><b>uncertainty</b></span><span style="margin:0px"><span style="margin:0px"> </span></span><span style="margin:0px">information for the problems at hand. Finally, I will open a window into the future technologies and applications such approach can enable. </span></span></span></span></font></span></span></span></span></span></div><font face="arial, sans-serif"><br></font></div><div style="color:rgb(0,0,0)"><font face="arial, sans-serif"><b><span style="margin:0px"><span style="margin:0px">BIO</span></span></b>: <span style="margin:0px;color:rgb(102,102,102)">Tolga Birdal is a Postdoctoral Research Fellow at Stanford University. He carries his research within the Geometric Computing Group of Prof. Leonidas Guibas. Previously, Tolga has defended his PhD thesis at the Computer Vision Group, Chair for Computer Aided Medical Procedures, Technical University of Munich led by Prof. Nassir Navab. He was also a Doktorand at Siemens AG. Tolga completed his Bachelors as an Electronics Engineer at 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 focused his research and development on large object detection, pose estimation and reconstruction using point clouds. Tolga 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 NeurIPS, CVPR, ICCV, ECCV, 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 <a href="http://tbirdal.me/" target="_blank">tbirdal.me</a>, </span><span style="margin:0px;color:rgb(102,102,102)"><a href="https://profiles.stanford.edu/tolga-birdal" target="_blank">https://profiles.stanford.edu/tolga-birdal</a></span><span style="margin:0px;color:rgb(102,102,102)">.</span></font></div><div id="gmail-m_-931738937922747375gmail-m_6606618964153194863gmail-m_6013610420488477655gmail-m_-6312913977678378189gmail-m_2026403581502646114appendonsend" style="font-size:small"></div><br></div><div style="font-size:small"><p style="color:rgb(60,64,67);letter-spacing:0.2px;white-space:pre-wrap"><b style="font-family:arial,sans-serif;letter-spacing:normal;color:rgb(34,34,34)">Host:</b><span style="font-family:arial,sans-serif;letter-spacing:normal;color:rgb(34,34,34)"> <a href="mailto:mwalter@ttic.edu" target="_blank">Matthew Walter</a></span><br></p><p style="font-size:14px;color:rgb(60,64,67);font-family:Roboto,Arial,sans-serif;letter-spacing:0.2px;white-space:pre-wrap"><br></p></div></div><div><div dir="ltr"><div dir="ltr"><div><div dir="ltr"><div><div dir="ltr"><div><div dir="ltr"><div><div dir="ltr"><div><div dir="ltr"><div><div dir="ltr"><font face="arial, helvetica, sans-serif">Mary C. Marre</font><div><font face="arial, helvetica, sans-serif">Faculty Administrative Support</font></div><div><i><font face="arial, helvetica, sans-serif" color="#3d85c6"><b>Toyota Technological Institute</b></font></i></div><div><i><font face="arial, helvetica, sans-serif" color="#3d85c6">6045 S. Kenwood Avenue</font></i></div><div><i><font face="arial, helvetica, sans-serif" color="#3d85c6">Room 517</font></i></div><div><i><font face="arial, helvetica, sans-serif" color="#3d85c6">Chicago, IL  60637</font></i></div><div><i><font face="arial, helvetica, sans-serif">p:(773) 834-1757</font></i></div><div><i><font face="arial, helvetica, sans-serif">f: (773) 357-6970</font></i></div><div><b><i><a href="mailto:mmarre@ttic.edu" target="_blank"><font face="arial, helvetica, sans-serif">mmarre@ttic.edu</font></a></i></b></div></div></div></div></div></div></div></div></div></div></div></div></div></div></div></div></div>
</blockquote></div></div>
</blockquote></div></div>