{"id":171004,"date":"2012-07-25T01:35:22","date_gmt":"2012-07-25T01:35:22","guid":{"rendered":"https:\/\/www.microsoft.com\/en-us\/research\/project\/decision-forests\/"},"modified":"2017-06-06T12:09:49","modified_gmt":"2017-06-06T19:09:49","slug":"decision-forests","status":"publish","type":"msr-project","link":"https:\/\/www.microsoft.com\/en-us\/research\/project\/decision-forests\/","title":{"rendered":"Decision Forests"},"content":{"rendered":"<table style=\"height: 353px; background-color: #dbd9d9;\" width=\"652\">\n<tbody>\n<tr>\n<td>\n<h2 align=\"center\">Decision Forests<\/h2>\n<h2 align=\"center\">for Computer Vision and<\/h2>\n<h2 align=\"center\">Medical Image Analysis<\/h2>\n<p align=\"center\"><strong>A. Criminisi and J. Shotton<\/strong><\/p>\n<p align=\"center\">Springer 2013, XIX, 368 p. 143 illus., 136 in color.<\/p>\n<p align=\"center\">ISBN 978-1-4471-4929-3<\/p>\n<\/td>\n<td>\u00a0<a class=\"msr-external-link glyph-append glyph-append-open-in-new-tab glyph-append-xsmall\" href=\"http:\/\/www.Alpha XR\/gp\/product\/1447149289\/ref=s9_simh_gw_p14_d0_i1?pf_rd_m=ATVPDKIKX0DER&pf_rd_s=center-2&pf_rd_r=0J09BB5MWBYTJP44VKH5&pf_rd_t=101&pf_rd_p=1389517282&pf_rd_i=507846\"><img loading=\"lazy\" decoding=\"async\" class=\"size-medium wp-image-214678 aligncenter\" src=\"https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2012\/07\/decisionforests_book_sm-200x300.jpg\" alt=\"decisionforests_book_sm\" width=\"200\" height=\"300\" srcset=\"https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2012\/07\/decisionforests_book_sm-200x300.jpg 200w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2012\/07\/decisionforests_book_sm.jpg 590w\" sizes=\"(max-width: 200px) 100vw, 200px\" \/><span class=\"sr-only\"> (opens in new tab)<\/span><\/a><\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n\t\t\t<div class=\"moray-accordion\" data-show=\"false\">\n\t\t\t\t\t\t\t<div class=\"clearfix\">\n\t\t\t\t\t<div\n\t\t\t\t\t\tclass=\"btn-group align-items-center mb-g float-sm-right\"\n\t\t\t\t\t\trole=\"group\"\n\t\t\t\t\t\tdata-mount=\"collapse-controls\"\n\t\t\t\t\t\tdata-bi-aN=\"accordion-collapse-controls\"\n\t\t\t\t\t\tdata-target=\"#accordion-1\">\n\t\t\t\t\t\t<button class=\"btn btn-link m-0\" aria-disabled=\"false\" aria-pressed=\"false\" data-action=\"expand\" type=\"button\" data-bi-cN=\"Expand all\">Expand all<\/button>\n\t\t\t\t\t\t<span aria-hidden=\"true\"> | <\/span>\n\t\t\t\t\t\t<button class=\"btn btn-link m-0\" aria-disabled=\"false\" aria-pressed=\"false\" data-action=\"collapse\" type=\"button\" data-bi-cN=\"Collapse all\">Collapse all<\/button>\n\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t\t\t<ul class=\"accordion\" id=\"accordion-1\">\n\t\t\t\t\t\t\t\t\t\t\t\t\t<li class=\"m-0\">\n\t\t\t<h4 class=\"accordion-header\">\n\t\t\t\t<button\n\t\t\t\t\tclass=\"btn btn-collapse m-0\"\n\t\t\t\t\ttype=\"button\"\n\t\t\t\t\tdata-mount=\"collapse\"\n\t\t\t\t\tdata-target=\"#collapse-2\"\n\t\t\t\t\taria-expanded=\"false\"\n\t\t\t\t\taria-controls=\"collapse-2\">\n\t\tBook Overview\t\t\t\t<\/button>\n\t\t\t<\/h4>\n\t\t\t<div class=\"collapse\" id=\"collapse-2\">\n\t\t\t\t<div class=\"accordion-body\">\n\t\t\t\t\t\t\t\t\t\t<p>This book presents a unified, efficient model of decision forests which can be used in a number of applications such as scene recognition from photographs, object recognition in images and automatic diagnosis from radiological scans. Such applications have traditionally been addressed by different, supervised or unsupervised machine learning techniques.<\/p>\n<p>However, in this book, diverse learning tasks including regression, classification and semi-supervised learning are all seen as instances of the same general decision forest model. The unified framework further extends to novel uses of forests in tasks such as density estimation and manifold learning. This unification carries both theoretical and practical advantages. For instance, the underlying single model gives us the opportunity to implement and optimize the general algorithm for all these tasks only once, and then easily adapt it to individual applications with relatively small changes.<\/p>\n<p><strong>Part I<\/strong> describes the general forest model which unifies classification, regression, density estimation, manifold learning, semi-supervised learning and active learning under the same flexible framework. The proposed model may be used both in a discriminative or generative way and may be applied to discrete or continuous, labelled or unlabelled data. It is based on a conventional training-testing framework, with the training phase trying to optimize a well defined energy function. Tasks such as classification or density estimation, supervised or unsupervised problems can all be addressed by setting a specific model for the objective function as well as the output prediction function.<\/p>\n<p><strong>Part II<\/strong> is a collection of invited chapters. Here various researchers show how it is possible to build different applications on top of the general forest model. Kinect-based player segmentation, semantic segmentation of photographs and automatic diagnosis of brain lesions are amongst the many applications discussed here.<\/p>\n<p><strong>Part III<\/strong> presents implementation details, documentation for the provided research software library, and some concluding remarks.<\/p>\n\t\t\t\t<\/div>\n\t\t\t<\/div>\n\t\t<\/li>\n\t\t\t\t\t\t<li class=\"m-0\">\n\t\t\t<h4 class=\"accordion-header\">\n\t\t\t\t<button\n\t\t\t\t\tclass=\"btn btn-collapse m-0\"\n\t\t\t\t\ttype=\"button\"\n\t\t\t\t\tdata-mount=\"collapse\"\n\t\t\t\t\tdata-target=\"#collapse-3\"\n\t\t\t\t\taria-expanded=\"false\"\n\t\t\t\t\taria-controls=\"collapse-3\">\n\t\tSolutions to exercises in the book\t\t\t\t<\/button>\n\t\t\t<\/h4>\n\t\t\t<div class=\"collapse\" id=\"collapse-3\">\n\t\t\t\t<div class=\"accordion-body\">\n\t\t\t\t\t\t\t\t\t\t<p style=\"text-align: center;\">Decision Forests\u00a0for Computer Vision and Medical Image Analysis<\/p>\n<p style=\"text-align: center;\">A. Criminisi and J. Shotton<\/p>\n<p style=\"text-align: center;\">Springer 2013<\/p>\n<h2>Chapter 4: Classification Forests<\/h2>\n<h3>Exercise 4.1<\/h3>\n<p><img loading=\"lazy\" decoding=\"async\" class=\"alignnone wp-image-180888 size-full\" src=\"https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2016\/02\/decisionforests-ex.4.1.a.png\" alt=\"ex.4.1.a.png\" width=\"300\" height=\"300\" srcset=\"https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2016\/02\/decisionforests-ex.4.1.a.png 300w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2016\/02\/decisionforests-ex.4.1.a-150x150.png 150w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2016\/02\/decisionforests-ex.4.1.a-180x180.png 180w\" sizes=\"(max-width: 300px) 100vw, 300px\" \/>\u00a0<img loading=\"lazy\" decoding=\"async\" class=\"alignnone wp-image-180889 size-full\" src=\"https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2016\/02\/decisionforests-ex.4.1.b.png\" alt=\"ex.4.1.b.png\" width=\"300\" height=\"300\" srcset=\"https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2016\/02\/decisionforests-ex.4.1.b.png 300w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2016\/02\/decisionforests-ex.4.1.b-150x150.png 150w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2016\/02\/decisionforests-ex.4.1.b-180x180.png 180w\" sizes=\"(max-width: 300px) 100vw, 300px\" \/><\/p>\n<p>Using many trees and linear splits reduces artifacts.<\/p>\n<h3>Exercise 4.2<\/h3>\n<p><img loading=\"lazy\" decoding=\"async\" class=\"alignnone size-full wp-image-180890\" src=\"https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2016\/02\/decisionforests-ex.4.2.a.png\" alt=\"ex.4.2.a.png\" width=\"300\" height=\"300\" srcset=\"https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2016\/02\/decisionforests-ex.4.2.a.png 300w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2016\/02\/decisionforests-ex.4.2.a-150x150.png 150w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2016\/02\/decisionforests-ex.4.2.a-180x180.png 180w\" sizes=\"(max-width: 300px) 100vw, 300px\" \/>\u00a0<img loading=\"lazy\" decoding=\"async\" class=\"alignnone size-full wp-image-180891\" src=\"https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2016\/02\/decisionforests-ex.4.2.b.png\" alt=\"ex.4.2.b.png\" width=\"300\" height=\"300\" srcset=\"https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2016\/02\/decisionforests-ex.4.2.b.png 300w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2016\/02\/decisionforests-ex.4.2.b-150x150.png 150w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2016\/02\/decisionforests-ex.4.2.b-180x180.png 180w\" sizes=\"(max-width: 300px) 100vw, 300px\" \/><\/p>\n<p>The quality of the uncertainty away from training data is affected by the type of split function (weak learner).<\/p>\n<h3>Exercise 4.3<\/h3>\n<p><img loading=\"lazy\" decoding=\"async\" class=\"alignnone size-full wp-image-180892\" src=\"https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2016\/02\/decisionforests-ex.4.3.a.png\" alt=\"ex.4.3.a.png\" width=\"300\" height=\"300\" srcset=\"https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2016\/02\/decisionforests-ex.4.3.a.png 300w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2016\/02\/decisionforests-ex.4.3.a-150x150.png 150w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2016\/02\/decisionforests-ex.4.3.a-180x180.png 180w\" sizes=\"(max-width: 300px) 100vw, 300px\" \/>\u00a0<img loading=\"lazy\" decoding=\"async\" class=\"alignnone size-full wp-image-180893\" src=\"https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2016\/02\/decisionforests-ex.4.3.b.png\" alt=\"ex.4.3.b.png\" width=\"300\" height=\"300\" srcset=\"https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2016\/02\/decisionforests-ex.4.3.b.png 300w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2016\/02\/decisionforests-ex.4.3.b-150x150.png 150w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2016\/02\/decisionforests-ex.4.3.b-180x180.png 180w\" sizes=\"(max-width: 300px) 100vw, 300px\" \/><\/p>\n<p>Using linear splits produces a possibly better separating surfaces.<\/p>\n<h3>Exercise 4.4<\/h3>\n<p><img loading=\"lazy\" decoding=\"async\" class=\"alignnone size-full wp-image-180894\" src=\"https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2016\/02\/decisionforests-ex.4.4.a.png\" alt=\"ex.4.4.a.png\" width=\"300\" height=\"300\" srcset=\"https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2016\/02\/decisionforests-ex.4.4.a.png 300w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2016\/02\/decisionforests-ex.4.4.a-150x150.png 150w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2016\/02\/decisionforests-ex.4.4.a-180x180.png 180w\" sizes=\"(max-width: 300px) 100vw, 300px\" \/>\u00a0<img loading=\"lazy\" decoding=\"async\" class=\"alignnone size-full wp-image-180895\" src=\"https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2016\/02\/decisionforests-ex.4.4.b.png\" alt=\"ex.4.4.b.png\" width=\"300\" height=\"300\" srcset=\"https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2016\/02\/decisionforests-ex.4.4.b.png 300w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2016\/02\/decisionforests-ex.4.4.b-150x150.png 150w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2016\/02\/decisionforests-ex.4.4.b-180x180.png 180w\" sizes=\"(max-width: 300px) 100vw, 300px\" \/><\/p>\n<p>Reducing the tree depth may cause underfitting and lower confidence.<\/p>\n<h3>Exercise 4.5<\/h3>\n<p><img loading=\"lazy\" decoding=\"async\" class=\"alignnone size-full wp-image-180896\" src=\"https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2016\/02\/decisionforests-ex.4.5.a.png\" alt=\"ex.4.5.a.png\" width=\"300\" height=\"300\" srcset=\"https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2016\/02\/decisionforests-ex.4.5.a.png 300w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2016\/02\/decisionforests-ex.4.5.a-150x150.png 150w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2016\/02\/decisionforests-ex.4.5.a-180x180.png 180w\" sizes=\"(max-width: 300px) 100vw, 300px\" \/>\u00a0<img loading=\"lazy\" decoding=\"async\" class=\"alignnone size-full wp-image-180897\" src=\"https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2016\/02\/decisionforests-ex.4.5.b.png\" alt=\"ex.4.5.b.png\" width=\"300\" height=\"300\" srcset=\"https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2016\/02\/decisionforests-ex.4.5.b.png 300w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2016\/02\/decisionforests-ex.4.5.b-150x150.png 150w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2016\/02\/decisionforests-ex.4.5.b-180x180.png 180w\" sizes=\"(max-width: 300px) 100vw, 300px\" \/><\/p>\n<p>Increasing\u00a0randomness may reduce overall prediction confidence.<\/p>\n<h2>Chapter 5: Regression Forests<\/h2>\n<h3>Exercise 5.1<\/h3>\n<p><img loading=\"lazy\" decoding=\"async\" class=\"alignnone size-full wp-image-180898\" src=\"https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2016\/02\/decisionforests-ex.5.1.a.png\" alt=\"ex.5.1.a.png\" width=\"300\" height=\"300\" srcset=\"https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2016\/02\/decisionforests-ex.5.1.a.png 300w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2016\/02\/decisionforests-ex.5.1.a-150x150.png 150w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2016\/02\/decisionforests-ex.5.1.a-180x180.png 180w\" sizes=\"(max-width: 300px) 100vw, 300px\" \/>\u00a0<img loading=\"lazy\" decoding=\"async\" class=\"alignnone size-full wp-image-180899\" src=\"https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2016\/02\/decisionforests-ex.5.1.b.png\" alt=\"ex.5.1.b.png\" width=\"300\" height=\"300\" srcset=\"https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2016\/02\/decisionforests-ex.5.1.b.png 300w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2016\/02\/decisionforests-ex.5.1.b-150x150.png 150w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2016\/02\/decisionforests-ex.5.1.b-180x180.png 180w\" sizes=\"(max-width: 300px) 100vw, 300px\" \/><\/p>\n<p>Large tree depth may lead to overfitting.<\/p>\n<h3>Exercise 5.2<\/h3>\n<p><img loading=\"lazy\" decoding=\"async\" class=\"alignnone size-full wp-image-180900\" src=\"https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2016\/02\/decisionforests-ex.5.2.a.png\" alt=\"ex.5.2.a.png\" width=\"300\" height=\"300\" srcset=\"https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2016\/02\/decisionforests-ex.5.2.a.png 300w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2016\/02\/decisionforests-ex.5.2.a-150x150.png 150w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2016\/02\/decisionforests-ex.5.2.a-180x180.png 180w\" sizes=\"(max-width: 300px) 100vw, 300px\" \/>\u00a0<img loading=\"lazy\" decoding=\"async\" class=\"alignnone size-full wp-image-180901\" src=\"https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2016\/02\/decisionforests-ex.5.2.b.png\" alt=\"ex.5.2.b.png\" width=\"300\" height=\"300\" srcset=\"https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2016\/02\/decisionforests-ex.5.2.b.png 300w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2016\/02\/decisionforests-ex.5.2.b-150x150.png 150w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2016\/02\/decisionforests-ex.5.2.b-180x180.png 180w\" sizes=\"(max-width: 300px) 100vw, 300px\" \/><\/p>\n<p>Larger\u00a0training noise yields larger\u00a0prediction uncertainty (wider pink region).<\/p>\n<h3>Exercise 5.3<\/h3>\n<p><img loading=\"lazy\" decoding=\"async\" class=\"alignnone size-full wp-image-180902\" src=\"https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2016\/02\/decisionforests-ex.5.3.a.png\" alt=\"ex.5.3.a.png\" width=\"300\" height=\"300\" srcset=\"https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2016\/02\/decisionforests-ex.5.3.a.png 300w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2016\/02\/decisionforests-ex.5.3.a-150x150.png 150w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2016\/02\/decisionforests-ex.5.3.a-180x180.png 180w\" sizes=\"(max-width: 300px) 100vw, 300px\" \/>\u00a0<img loading=\"lazy\" decoding=\"async\" class=\"alignnone size-full wp-image-180903\" src=\"https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2016\/02\/decisionforests-ex.5.3.b.png\" alt=\"ex.5.3.b.png\" width=\"300\" height=\"300\" srcset=\"https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2016\/02\/decisionforests-ex.5.3.b.png 300w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2016\/02\/decisionforests-ex.5.3.b-150x150.png 150w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2016\/02\/decisionforests-ex.5.3.b-180x180.png 180w\" sizes=\"(max-width: 300px) 100vw, 300px\" \/><\/p>\n<p><img loading=\"lazy\" decoding=\"async\" class=\"alignnone size-full wp-image-180904\" src=\"https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2016\/02\/decisionforests-ex.5.3.c.png\" alt=\"ex.5.3.c.png\" width=\"300\" height=\"300\" srcset=\"https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2016\/02\/decisionforests-ex.5.3.c.png 300w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2016\/02\/decisionforests-ex.5.3.c-150x150.png 150w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2016\/02\/decisionforests-ex.5.3.c-180x180.png 180w\" sizes=\"(max-width: 300px) 100vw, 300px\" \/>\u00a0<img loading=\"lazy\" decoding=\"async\" class=\"alignnone size-full wp-image-180905\" src=\"https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2016\/02\/decisionforests-ex.5.3.d.png\" alt=\"ex.5.3.d.png\" width=\"300\" height=\"300\" srcset=\"https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2016\/02\/decisionforests-ex.5.3.d.png 300w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2016\/02\/decisionforests-ex.5.3.d-150x150.png 150w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2016\/02\/decisionforests-ex.5.3.d-180x180.png 180w\" sizes=\"(max-width: 300px) 100vw, 300px\" \/><\/p>\n<p>Non-linear curve fitting in diverse examples. Note the relatively smooth interpolation and extrapolation behaviour.<\/p>\n<h3>Exercise 5.4<\/h3>\n<p><img loading=\"lazy\" decoding=\"async\" class=\"alignnone size-full wp-image-180906\" src=\"https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2016\/02\/decisionforests-ex.5.4.a.png\" alt=\"ex.5.4.a.png\" width=\"300\" height=\"300\" srcset=\"https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2016\/02\/decisionforests-ex.5.4.a.png 300w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2016\/02\/decisionforests-ex.5.4.a-150x150.png 150w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2016\/02\/decisionforests-ex.5.4.a-180x180.png 180w\" sizes=\"(max-width: 300px) 100vw, 300px\" \/>\u00a0<img loading=\"lazy\" decoding=\"async\" class=\"alignnone size-full wp-image-180907\" src=\"https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2016\/02\/decisionforests-ex.5.4.b.png\" alt=\"ex.5.4.b.png\" width=\"300\" height=\"300\" srcset=\"https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2016\/02\/decisionforests-ex.5.4.b.png 300w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2016\/02\/decisionforests-ex.5.4.b-150x150.png 150w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2016\/02\/decisionforests-ex.5.4.b-180x180.png 180w\" sizes=\"(max-width: 300px) 100vw, 300px\" \/><\/p>\n<p>Single function\u00a0regression does not capture the inherently ambiguous central region. But at least it\u00a0returns an associated high uncertainty.<\/p>\n<h2>Chapter 6: Decision Forests<\/h2>\n<h3>Exercise 6.1<\/h3>\n<p><img loading=\"lazy\" decoding=\"async\" class=\"alignnone size-medium wp-image-180909\" src=\"https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2016\/02\/decisionforests-ex.6.1.a-300x300.png\" alt=\"ex.6.1.a.png\" width=\"300\" height=\"300\" srcset=\"https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2016\/02\/decisionforests-ex.6.1.a-300x300.png 300w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2016\/02\/decisionforests-ex.6.1.a-150x150.png 150w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2016\/02\/decisionforests-ex.6.1.a-180x180.png 180w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2016\/02\/decisionforests-ex.6.1.a.png 320w\" sizes=\"(max-width: 300px) 100vw, 300px\" \/>\u00a0<img loading=\"lazy\" decoding=\"async\" class=\"alignnone size-medium wp-image-180908\" src=\"https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2016\/02\/decisionforests-ex.6.1.b-300x300.png\" alt=\"ex.6.1.b.png\" width=\"300\" height=\"300\" srcset=\"https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2016\/02\/decisionforests-ex.6.1.b-300x300.png 300w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2016\/02\/decisionforests-ex.6.1.b-150x150.png 150w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2016\/02\/decisionforests-ex.6.1.b-180x180.png 180w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2016\/02\/decisionforests-ex.6.1.b.png 320w\" sizes=\"(max-width: 300px) 100vw, 300px\" \/><\/p>\n<p>Too deep trees may\u00a0cause overfitting.<\/p>\n<h3>Exercise 6.2<\/h3>\n<p><img loading=\"lazy\" decoding=\"async\" class=\"alignnone size-medium wp-image-180910\" src=\"https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2016\/02\/decisionforests-ex.6.2.a-300x300.png\" alt=\"ex.6.2.a.png\" width=\"300\" height=\"300\" srcset=\"https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2016\/02\/decisionforests-ex.6.2.a-300x300.png 300w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2016\/02\/decisionforests-ex.6.2.a-150x150.png 150w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2016\/02\/decisionforests-ex.6.2.a-180x180.png 180w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2016\/02\/decisionforests-ex.6.2.a.png 320w\" sizes=\"(max-width: 300px) 100vw, 300px\" \/>\u00a0<img loading=\"lazy\" decoding=\"async\" class=\"alignnone size-medium wp-image-180911\" src=\"https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2016\/02\/decisionforests-ex.6.2.b-300x300.png\" alt=\"ex.6.2.b.png\" width=\"300\" height=\"300\" srcset=\"https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2016\/02\/decisionforests-ex.6.2.b-300x300.png 300w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2016\/02\/decisionforests-ex.6.2.b-150x150.png 150w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2016\/02\/decisionforests-ex.6.2.b-180x180.png 180w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2016\/02\/decisionforests-ex.6.2.b.png 320w\" sizes=\"(max-width: 300px) 100vw, 300px\" \/><\/p>\n<p>Too deep trees may cause overfitting.<\/p>\n<h3>Exercise 6.3<\/h3>\n<p><img loading=\"lazy\" decoding=\"async\" class=\"alignnone size-medium wp-image-180913\" src=\"https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2016\/02\/decisionforests-ex.6.3.a-300x300.png\" alt=\"ex.6.3.a.png\" width=\"300\" height=\"300\" srcset=\"https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2016\/02\/decisionforests-ex.6.3.a-300x300.png 300w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2016\/02\/decisionforests-ex.6.3.a-150x150.png 150w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2016\/02\/decisionforests-ex.6.3.a-180x180.png 180w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2016\/02\/decisionforests-ex.6.3.a.png 320w\" sizes=\"(max-width: 300px) 100vw, 300px\" \/>\u00a0<img loading=\"lazy\" decoding=\"async\" class=\"alignnone size-medium wp-image-180912\" src=\"https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2016\/02\/decisionforests-ex.6.3.b-300x300.png\" alt=\"ex.6.3.b.png\" width=\"300\" height=\"300\" srcset=\"https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2016\/02\/decisionforests-ex.6.3.b-300x300.png 300w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2016\/02\/decisionforests-ex.6.3.b-150x150.png 150w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2016\/02\/decisionforests-ex.6.3.b-180x180.png 180w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2016\/02\/decisionforests-ex.6.3.b.png 320w\" sizes=\"(max-width: 300px) 100vw, 300px\" \/><\/p>\n<p>Too deep trees may cause overfitting.<\/p>\n<h3>Exercise 6.4<\/h3>\n<p><img loading=\"lazy\" decoding=\"async\" class=\"alignnone size-medium wp-image-180914\" src=\"https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2016\/02\/decisionforests-ex.6.5.a-300x300.png\" alt=\"ex.6.5.a.png\" width=\"300\" height=\"300\" srcset=\"https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2016\/02\/decisionforests-ex.6.5.a-300x300.png 300w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2016\/02\/decisionforests-ex.6.5.a-150x150.png 150w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2016\/02\/decisionforests-ex.6.5.a-180x180.png 180w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2016\/02\/decisionforests-ex.6.5.a.png 320w\" sizes=\"(max-width: 300px) 100vw, 300px\" \/>\u00a0<img loading=\"lazy\" decoding=\"async\" class=\"alignnone size-medium wp-image-180915\" src=\"https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2016\/02\/decisionforests-ex.6.5.b-300x300.png\" alt=\"ex.6.5.b.png\" width=\"300\" height=\"300\" srcset=\"https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2016\/02\/decisionforests-ex.6.5.b-300x300.png 300w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2016\/02\/decisionforests-ex.6.5.b-150x150.png 150w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2016\/02\/decisionforests-ex.6.5.b-180x180.png 180w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2016\/02\/decisionforests-ex.6.5.b.png 320w\" sizes=\"(max-width: 300px) 100vw, 300px\" \/><\/p>\n<p>Too deep trees may cause overfitting.\u00a0Some of the visible streaky artifacts are due to the use of axis-aligned weak learners.<\/p>\n<h2>Chapter 8: Semi-supervised Classification Forests<\/h2>\n<h3>Exercise 8.1<\/h3>\n<p><img loading=\"lazy\" decoding=\"async\" class=\"alignnone size-medium wp-image-180917\" src=\"https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2016\/02\/decisionforests-ex.8.1.a-300x300.png\" alt=\"ex.8.1.a.png\" width=\"300\" height=\"300\" srcset=\"https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2016\/02\/decisionforests-ex.8.1.a-300x300.png 300w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2016\/02\/decisionforests-ex.8.1.a-150x150.png 150w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2016\/02\/decisionforests-ex.8.1.a-180x180.png 180w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2016\/02\/decisionforests-ex.8.1.a.png 320w\" sizes=\"(max-width: 300px) 100vw, 300px\" \/>\u00a0<img loading=\"lazy\" decoding=\"async\" class=\"alignnone size-medium wp-image-180916\" src=\"https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2016\/02\/decisionforests-ex.8.1.b-300x300.png\" alt=\"ex.8.1.b.png\" width=\"300\" height=\"300\" srcset=\"https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2016\/02\/decisionforests-ex.8.1.b-300x300.png 300w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2016\/02\/decisionforests-ex.8.1.b-150x150.png 150w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2016\/02\/decisionforests-ex.8.1.b-180x180.png 180w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2016\/02\/decisionforests-ex.8.1.b.png 320w\" sizes=\"(max-width: 300px) 100vw, 300px\" \/><\/p>\n<p>Note the larger uncertainty in the central region (left image). A single tree is always over-confident.<\/p>\n<h3>Exercise 8.2<\/h3>\n<p><img loading=\"lazy\" decoding=\"async\" class=\"alignnone size-medium wp-image-180919\" src=\"https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2016\/02\/decisionforests-ex.8.2.a-300x300.png\" alt=\"ex.8.2.a.png\" width=\"300\" height=\"300\" srcset=\"https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2016\/02\/decisionforests-ex.8.2.a-300x300.png 300w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2016\/02\/decisionforests-ex.8.2.a-150x150.png 150w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2016\/02\/decisionforests-ex.8.2.a-180x180.png 180w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2016\/02\/decisionforests-ex.8.2.a.png 320w\" sizes=\"(max-width: 300px) 100vw, 300px\" \/>\u00a0<img loading=\"lazy\" decoding=\"async\" class=\"alignnone size-medium wp-image-180918\" src=\"https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2016\/02\/decisionforests-ex.8.2.b-300x300.png\" alt=\"ex.8.2.b.png\" width=\"300\" height=\"300\" srcset=\"https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2016\/02\/decisionforests-ex.8.2.b-300x300.png 300w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2016\/02\/decisionforests-ex.8.2.b-150x150.png 150w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2016\/02\/decisionforests-ex.8.2.b-180x180.png 180w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2016\/02\/decisionforests-ex.8.2.b.png 320w\" sizes=\"(max-width: 300px) 100vw, 300px\" \/><\/p>\n<p>Adding further supervised data in the central region helps increase the\u00a0prediction\u00a0confidence.<\/p>\n<h3>Exercise 8.3<\/h3>\n<p><img loading=\"lazy\" decoding=\"async\" class=\"alignnone size-medium wp-image-180921\" src=\"https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2016\/02\/decisionforests-ex.8.3.a-300x300.png\" alt=\"ex.8.3.a.png\" width=\"300\" height=\"300\" srcset=\"https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2016\/02\/decisionforests-ex.8.3.a-300x300.png 300w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2016\/02\/decisionforests-ex.8.3.a-150x150.png 150w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2016\/02\/decisionforests-ex.8.3.a-180x180.png 180w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2016\/02\/decisionforests-ex.8.3.a.png 320w\" sizes=\"(max-width: 300px) 100vw, 300px\" \/>\u00a0<img loading=\"lazy\" decoding=\"async\" class=\"alignnone size-medium wp-image-180920\" src=\"https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2016\/02\/decisionforests-ex.8.3.b-300x300.png\" alt=\"ex.8.3.b.png\" width=\"300\" height=\"300\" srcset=\"https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2016\/02\/decisionforests-ex.8.3.b-300x300.png 300w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2016\/02\/decisionforests-ex.8.3.b-150x150.png 150w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2016\/02\/decisionforests-ex.8.3.b-180x180.png 180w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2016\/02\/decisionforests-ex.8.3.b.png 320w\" sizes=\"(max-width: 300px) 100vw, 300px\" \/><\/p>\n<p>Confidence decreases with training noise and increases with tree depth.<\/p>\n<h3>Exercise 8.4<\/h3>\n<p><img loading=\"lazy\" decoding=\"async\" class=\"alignnone size-medium wp-image-180923\" src=\"https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2016\/02\/decisionforests-ex.8.4.a-300x300.png\" alt=\"ex.8.4.a.png\" width=\"300\" height=\"300\" srcset=\"https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2016\/02\/decisionforests-ex.8.4.a-300x300.png 300w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2016\/02\/decisionforests-ex.8.4.a-150x150.png 150w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2016\/02\/decisionforests-ex.8.4.a-180x180.png 180w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2016\/02\/decisionforests-ex.8.4.a.png 320w\" sizes=\"(max-width: 300px) 100vw, 300px\" \/>\u00a0<img loading=\"lazy\" decoding=\"async\" class=\"alignnone size-medium wp-image-180922\" src=\"https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2016\/02\/decisionforests-ex.8.4.b-300x300.png\" alt=\"ex.8.4.b.png\" width=\"300\" height=\"300\" srcset=\"https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2016\/02\/decisionforests-ex.8.4.b-300x300.png 300w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2016\/02\/decisionforests-ex.8.4.b-150x150.png 150w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2016\/02\/decisionforests-ex.8.4.b-180x180.png 180w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2016\/02\/decisionforests-ex.8.4.b.png 320w\" sizes=\"(max-width: 300px) 100vw, 300px\" \/><\/p>\n<p>Single trees are over-confident. Using many random forests produces smooth uncertainty in the transition regions.<\/p>\n<h3>Exercise 8.5<\/h3>\n<p><img loading=\"lazy\" decoding=\"async\" class=\"alignnone size-medium wp-image-180924\" src=\"https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2016\/02\/decisionforests-ex.8.5.a-300x300.png\" alt=\"ex.8.5.a.png\" width=\"300\" height=\"300\" srcset=\"https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2016\/02\/decisionforests-ex.8.5.a-300x300.png 300w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2016\/02\/decisionforests-ex.8.5.a-150x150.png 150w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2016\/02\/decisionforests-ex.8.5.a-180x180.png 180w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2016\/02\/decisionforests-ex.8.5.a.png 320w\" sizes=\"(max-width: 300px) 100vw, 300px\" \/>\u00a0<img loading=\"lazy\" decoding=\"async\" class=\"alignnone size-medium wp-image-180925\" src=\"https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2016\/02\/decisionforests-ex.8.5.b-300x300.png\" alt=\"ex.8.5.b.png\" width=\"300\" height=\"300\" srcset=\"https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2016\/02\/decisionforests-ex.8.5.b-300x300.png 300w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2016\/02\/decisionforests-ex.8.5.b-150x150.png 150w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2016\/02\/decisionforests-ex.8.5.b-180x180.png 180w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2016\/02\/decisionforests-ex.8.5.b.png 320w\" sizes=\"(max-width: 300px) 100vw, 300px\" \/><\/p>\n<p>Adding the amount of supervision in regions of low confidence increases the prediction accuracy and the overall confidence.<\/p>\n\t\t\t\t<\/div>\n\t\t\t<\/div>\n\t\t<\/li>\n\t\t\t\t\t\t<li class=\"m-0\">\n\t\t\t<h4 class=\"accordion-header\">\n\t\t\t\t<button\n\t\t\t\t\tclass=\"btn btn-collapse m-0\"\n\t\t\t\t\ttype=\"button\"\n\t\t\t\t\tdata-mount=\"collapse\"\n\t\t\t\t\tdata-target=\"#collapse-4\"\n\t\t\t\t\taria-expanded=\"false\"\n\t\t\t\t\taria-controls=\"collapse-4\">\n\t\tLinks\t\t\t\t<\/button>\n\t\t\t<\/h4>\n\t\t\t<div class=\"collapse\" id=\"collapse-4\">\n\t\t\t\t<div class=\"accordion-body\">\n\t\t\t\t\t\t\t\t\t\t<ul>\n<li><a class=\"msr-external-link glyph-append glyph-append-open-in-new-tab glyph-append-xsmall\" href=\"http:\/\/www.springer.com\/computer\/image+processing\/book\/978-1-4471-4928-6\">Order copy of book from Springer <span class=\"sr-only\"> (opens in new tab)<\/span><\/a><\/li>\n<li><a class=\"msr-external-link glyph-append glyph-append-open-in-new-tab glyph-append-xsmall\" href=\"http:\/\/www.Alpha XR\/Decision-Computer-Analysis-Advances-Recognition\/dp\/1447149289\/ref=sr_1_2?ie=UTF8&qid=1357811532&sr=8-2&keywords=criminisi+forest\">Order copy of book from Amazon<span class=\"sr-only\"> (opens in new tab)<\/span><\/a><\/li>\n<\/ul>\n\t\t\t\t<\/div>\n\t\t\t<\/div>\n\t\t<\/li>\n\t\t\t\t\t\t\t\t\t<\/ul>\n\t\t<\/div>\n\t\t\n","protected":false},"excerpt":{"rendered":"<p>Decision Forests for Computer Vision and Medical Image Analysis A. Criminisi and J. Shotton Springer 2013, XIX, 368 p. 143 illus., 136 in color. ISBN 978-1-4471-4929-3 \u00a0<\/p>\n","protected":false},"featured_media":0,"template":"","meta":{"msr-url-field":"","msr-podcast-episode":"","msrModifiedDate":"","msrModifiedDateEnabled":false,"ep_exclude_from_search":false,"footnotes":""},"research-area":[13556,13562],"msr-locale":[268875],"msr-impact-theme":[],"msr-pillar":[],"class_list":["post-171004","msr-project","type-msr-project","status-publish","hentry","msr-research-area-artificial-intelligence","msr-research-area-computer-vision","msr-locale-en_us","msr-archive-status-active"],"msr_project_start":"2012-07-25","related-publications":[160289,166740,162242,164123,160924,166741,162366,153608,164356,160987,167701,162557,157653,164357,160988,167702,162558,157665,164368,161084,167703,162773,157841,164375,161344,168300,162774,158288,164555,161346,215422,162775,158765,165516,161612,238041,163103,159746,166600,161784,163195,159916,166601,161953,163225,159935,166739,162240,163512],"related-downloads":[234973],"related-videos":[],"related-groups":[],"related-events":[],"related-opportunities":[],"related-posts":[467505],"related-articles":[],"tab-content":[],"slides":[],"related-researchers":[],"msr_research_lab":[],"msr_impact_theme":[],"_links":{"self":[{"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-project\/171004"}],"collection":[{"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-project"}],"about":[{"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/types\/msr-project"}],"version-history":[{"count":1,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-project\/171004\/revisions"}],"predecessor-version":[{"id":238589,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-project\/171004\/revisions\/238589"}],"wp:attachment":[{"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/media?parent=171004"}],"wp:term":[{"taxonomy":"msr-research-area","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/research-area?post=171004"},{"taxonomy":"msr-locale","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-locale?post=171004"},{"taxonomy":"msr-impact-theme","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-impact-theme?post=171004"},{"taxonomy":"msr-pillar","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-pillar?post=171004"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}