%PDF-1.3 1 0 obj << /Kids [ 4 0 R 5 0 R 6 0 R 7 0 R 8 0 R 9 0 R 10 0 R 11 0 R 12 0 R ] /Type /Pages /Count 9 >> endobj 2 0 obj << /Subject (Neural Information Processing Systems http\072\057\057nips\056cc\057) /Publisher (Curran Associates) /Language (en\055US) /Created (2010) /Description-Abstract (In many machine learning domains \050such as scene understanding\051\054 several related sub\055tasks \050such as scene categorization\054 depth estimation\054 object detection\051 operate on the same raw data and provide correlated outputs\056 Each of these tasks is often notoriously hard\054 and state\055of\055the\055art classifiers already exist for many sub\055tasks\056 It is desirable to have an algorithm that can capture such correlation without requiring to make any changes to the inner workings of any classifier\056 We propose Feedback Enabled Cascaded Classification Models \050FE\055CCM\051\054 that maximizes the joint likelihood of the sub\055tasks\054 while requiring only a \217black\055box\220 interface to the original classifier for each sub\055task\056 We use a two\055layer cascade of classifiers\054 which are repeated instantiations of the original ones\054 with the output of the first layer fed into the second layer as input\056 Our training method involves a feedback step that allows later classifiers to provide earlier classifiers information about what error modes to focus on\056 We show that our method significantly improves performance in all the sub\055tasks in two different domains\072 \050i\051 scene understanding\054 where we consider depth estimation\054 scene categorization\054 event categorization\054 object detection\054 geometric labeling and saliency detection\054 and \050ii\051 robotic grasping\054 where we consider grasp point detection and object classification\056) /Producer (Python PDF Library \055 http\072\057\057pybrary\056net\057pyPdf\057) /Title (Towards Holistic Scene Understanding\072 Feedback Enabled Cascaded Classification Models) /Date (2010) /Type (Conference Proceedings) /firstpage (1351) /Book (Advances in Neural Information Processing Systems 23) /Description (Paper accepted and presented at the Neural Information Processing Systems Conference \050http\072\057\057nips\056cc\057\051) /Editors (J\056D\056 Lafferty and C\056K\056I\056 Williams and J\056 Shawe\055Taylor and R\056S\056 Zemel and A\056 Culotta) /Author (Congcong Li\054 Adarsh Kowdle\054 Ashutosh Saxena\054 Tsuhan Chen) /lastpage (1359) >> endobj 3 0 obj << /Type /Catalog /Pages 1 0 R >> endobj 4 0 obj << /Contents 13 0 R /Parent 1 0 R /Type /Page /Resources 14 0 R /MediaBox [ 0 0 612 792 ] >> endobj 5 0 obj << /Contents 36 0 R /Parent 1 0 R /Type /Page /Resources 37 0 R /MediaBox [ 0 0 612 792 ] >> endobj 6 0 obj << /Contents 38 0 R /Parent 1 0 R /Type /Page /Resources 39 0 R /MediaBox [ 0 0 612 792 ] >> endobj 7 0 obj << /Contents 86 0 R /Parent 1 0 R /Type /Page /Resources 87 0 R /MediaBox [ 0 0 612 792 ] >> endobj 8 0 obj << /Contents 96 0 R /Parent 1 0 R /Type /Page /Resources 97 0 R /MediaBox [ 0 0 612 792 ] >> endobj 9 0 obj << /Contents 190 0 R /Parent 1 0 R /Type /Page /Resources 191 0 R /MediaBox [ 0 0 612 792 ] >> endobj 10 0 obj << /Contents 192 0 R /Parent 1 0 R /Type /Page /Resources 193 0 R /MediaBox [ 0 0 612 792 ] >> endobj 11 0 obj << /Contents 194 0 R /Parent 1 0 R /Type /Page /Resources 195 0 R /MediaBox [ 0 0 612 792 ] >> endobj 12 0 obj << /Contents 201 0 R /Parent 1 0 R /Type /Page /Resources 202 0 R /MediaBox [ 0 0 612 792 ] >> endobj 13 0 obj << /Length 3047 /Filter /FlateDecode >> stream x}YY~_P>6/Xo `x E$PxSWSD&p&+LMu~6_}z$ۄiy:l,"dYey7my6ejEiط{TwWkcMwn. ,6(]mӺޫWԾ5wT=Zem[D}[G<}zcy8lji%$`n~ɢ9Q2(X?s?biխv4#E}ѝ';\;RX?Q|8ctm<6,dIUgF!a2;d~SL9KHTmo9KVR?76_'3?) f*.NY+vYx{] I=0\8J"Naoa/JX;_]:5dIZ!?-H;V')bMv[PoD^\Llt0]xH;wx@̏Qw,44}M,-{}d?_ Ar4Y0