{"id":239528,"date":"2016-06-20T00:00:22","date_gmt":"2016-06-20T07:00:22","guid":{"rendered":"https:\/\/www.microsoft.com\/en-us\/research\/?post_type=msr-event&p=239528"},"modified":"2018-07-26T14:21:37","modified_gmt":"2018-07-26T21:21:37","slug":"fitting-surface-models-data-accuracy-speed-robustness","status":"publish","type":"msr-event","link":"https:\/\/www.microsoft.com\/en-us\/research\/event\/fitting-surface-models-data-accuracy-speed-robustness\/","title":{"rendered":"Fitting surface models to data: Accuracy, Speed, Robustness"},"content":{"rendered":"

Andrew Fitzgibbon, Microsoft
\nJon Taylor, PerceptiveIO<\/p>\n","protected":false},"excerpt":{"rendered":"

Fitting surface models to data: Accuracy, Speed, Robustness, CVPR Tutorial, Sunday June 26th, 2016. Led by Andrew Fitzgibbon, Microsoft and Jon Taylor, PerceptiveIO<\/p>\n","protected":false},"featured_media":0,"template":"","meta":{"msr-url-field":"","msr-podcast-episode":"","msrModifiedDate":"","msrModifiedDateEnabled":false,"ep_exclude_from_search":false,"msr_startdate":"2016-06-26","msr_enddate":"2016-06-26","msr_location":"CVPR Conference","msr_expirationdate":"2016-06-27","msr_event_recording_link":"","msr_event_link":"","msr_event_link_redirect":false,"msr_event_time":"09:00 - 17:15","msr_hide_region":false,"msr_private_event":false,"footnotes":""},"research-area":[13556],"msr-region":[239178],"msr-event-type":[197941],"msr-video-type":[],"msr-locale":[268875],"msr-program-audience":[],"msr-post-option":[],"msr-impact-theme":[],"class_list":["post-239528","msr-event","type-msr-event","status-publish","hentry","msr-research-area-artificial-intelligence","msr-region-europe","msr-event-type-conferences","msr-locale-en_us"],"msr_about":"Andrew Fitzgibbon, Microsoft\r\nJon Taylor, PerceptiveIO","tab-content":[{"id":0,"name":"About","content":"CVPR Tutorial<\/b><\/span>\r\n\r\nIn vision and machine learning, almost everything we do may be considered to be a form of model fitting. Whether estimating the parameters of a convolutional neural network, computing structure and motion from image collections, tracking objects in video, computing low-dimensional representations of datasets, estimating parameters for an inference model such as Markov random fields, or extracting shape spaces such as active appearance models, it almost always boils down to minimizing an objective containing some parameters of interest as well as some latent or nuisance parameters.<\/span>\u00a0 <\/span>This tutorial will describe several tools and techniques for solving such optimization problems, with a focus on fitting 3D smooth-surface models, such as subdivision surfaces, to 2D and 3D data.<\/span>\r\n\r\n \r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n
09:00<\/td>\r\nIntro: Applications in vision and graphics.\r\n\r\n\u00b7\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0 Lots of exciting and inspirational examples of model fitting:\r\n\r\no\u00a0\u00a0 Kinetre (Siggraph 12)\r\n\r\no\u00a0\u00a0 Dolphins (PAMI 13)\r\n\r\no\u00a0\u00a0 Nonrigid tracking (Siggraph 14)\r\n\r\no\u00a0\u00a0 FlexSense (CHI 15)\r\n\r\no\u00a0\u00a0 Hand tracking (Siggraph 15)\r\n\r\n\u00b7\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0 Preview of the day<\/td>\r\n<\/tr>\r\n
09:20<\/td>\r\nSession I: Matrix and vector calculus\r\n\r\n\u00b7\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0 vector functions and the Jacobian, generalized Jacobian\r\n\r\n\u00b7\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0 advanced matrix operations: block operations, kronecker products etc\r\n\r\n\u00b7\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0 derivatives of matrix expressions\r\n\r\n\u00b7\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0 finite-difference versus symbolic derivatives\r\n\r\n\u00b7\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0 sparse matrices and sparse storage\r\n\r\n\u00b7\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0 derivatives of minimization problems\r\n\r\n\u00b7\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0 nonlinear optimization, Gauss-Newton and Levenberg-Marquardt algorithms<\/td>\r\n<\/tr>\r\n
10:30<\/td>\r\nCoffee<\/td>\r\n<\/tr>\r\n
10:45<\/td>\r\nSession II: Curves, Surfaces, Correspondences\r\n\r\n\u00b7\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0 What is a surface?\u00a0 Parametric descriptions of curves and surfaces\r\n\r\n\u00b7\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0 Surfaces and data points: closest point operations\r\n\r\n\u00b7\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0 Fitting surfaces to data: correspondences\r\n\r\no\u00a0\u00a0 Iterated closest points\r\n\r\no\u00a0\u00a0 \"Lifting\" correspondences\r\n\r\n\u00b7\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0 Worked example: Gauss's Ceres problem<\/td>\r\n<\/tr>\r\n
11:40<\/td>\r\nBreak and stretch<\/td>\r\n<\/tr>\r\n
11:45<\/td>\r\nSession III: More on surfaces\r\n\r\n\u00b7\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0 Subdivision surfaces in 3D\r\n\r\n\u00b7\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0 Taking walks in parameter space\r\n\r\n\u00b7\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0 Implementing for speed<\/td>\r\n<\/tr>\r\n
12:30<\/td>\r\nLunch<\/td>\r\n<\/tr>\r\n
14:00<\/td>\r\nSession III: Robustness and speed\r\n\r\n\u00b7\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0 Using robust kernels in Levenberg-Marquardt (general interest)\r\n\r\no\u00a0\u00a0 A great example of where \"lifting\" really helps\r\n\r\n\u00b7\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0 Implementing rotations: quaternions vs infinitesimals with recentering\r\n\r\n\u00b7\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0 Schur complement QR<\/td>\r\n<\/tr>\r\n
15:00<\/td>\r\nCoffee\/Stretch<\/td>\r\n<\/tr>\r\n
15:15<\/td>\r\nSession IV: Software\r\n\r\n\u00b7\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0 OpenSubdiv\r\n\r\n\u00b7\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0 Eigen\r\n\r\n\u00b7\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0 Ceres\r\n\r\n\u00b7\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0 Opt\r\n\r\n\u00b7\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0 AD tools: Theano etc<\/td>\r\n<\/tr>\r\n
16:15<\/td>\r\nMore coffee, more stretching<\/td>\r\n<\/tr>\r\n
16:30<\/td>\r\nSession V: Conclusions and open problems\r\n\r\n\u00b7\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0 Topology adaptation\r\n\r\n\u00b7\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0 Where are<\/em> the local minima?\r\n\r\n\u00b7\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0 And where lifting really hurts: VarPro algorithms<\/td>\r\n<\/tr>\r\n
17:15<\/td>\r\nClose<\/td>\r\n<\/tr>\r\n<\/tbody>\r\n<\/table>"}],"msr_startdate":"2016-06-26","msr_enddate":"2016-06-26","msr_event_time":"09:00 - 17:15","msr_location":"CVPR Conference","msr_event_link":"","msr_event_recording_link":"","msr_startdate_formatted":"June 26, 2016","msr_register_text":"Watch now","msr_cta_link":"","msr_cta_text":"","msr_cta_bi_name":"","featured_image_thumbnail":null,"event_excerpt":"Fitting surface models to data: Accuracy, Speed, Robustness, CVPR Tutorial, Sunday June 26th, 2016. Led by Andrew Fitzgibbon, Microsoft and Jon Taylor, PerceptiveIO","msr_research_lab":[],"related-researchers":[],"msr_impact_theme":[],"related-academic-programs":[],"related-groups":[],"related-projects":[238767],"related-opportunities":[],"related-publications":[],"related-videos":[],"related-posts":[],"_links":{"self":[{"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-event\/239528"}],"collection":[{"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-event"}],"about":[{"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/types\/msr-event"}],"version-history":[{"count":2,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-event\/239528\/revisions"}],"predecessor-version":[{"id":497831,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-event\/239528\/revisions\/497831"}],"wp:attachment":[{"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/media?parent=239528"}],"wp:term":[{"taxonomy":"msr-research-area","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/research-area?post=239528"},{"taxonomy":"msr-region","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-region?post=239528"},{"taxonomy":"msr-event-type","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-event-type?post=239528"},{"taxonomy":"msr-video-type","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-video-type?post=239528"},{"taxonomy":"msr-locale","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-locale?post=239528"},{"taxonomy":"msr-program-audience","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-program-audience?post=239528"},{"taxonomy":"msr-post-option","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-post-option?post=239528"},{"taxonomy":"msr-impact-theme","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-impact-theme?post=239528"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}