280 likes | 494 Views
Models for Scene Understanding – Global Energy models and a Style-Parameterized boosting algorithm ( StyP -Boost) . Jonathan Warrell, 1 Simon Prince, 2 Philip Torr, 1 Lubor Ladicky, 1 Chris Russell 1 1 Oxford Brookes University, 2 University College London. Overview .
E N D
Models for Scene Understanding – Global Energy models and a Style-Parameterized boosting algorithm (StyP-Boost) Jonathan Warrell,1 Simon Prince,2 Philip Torr,1 Lubor Ladicky,1 Chris Russell1 1Oxford Brookes University, 2University College London
Overview • CRF-based semantic segmentation • Recent models • Detectors • Stereo • Co-occurence • Hierarchical Energies • Style parameterized boosting (StyP-Boost) • Open questions / problems
CRF-based semantic segmentation • Semantic segmentation = dense labeling using fixed object set
CRF-based semantic segmentation • Conditional Random Field model (pairwise) Observed Variables Hidden Variables Unary Pairwise
Example: -expansion Status: Tree Ground Initialize with Tree Expand Ground Expand House Expand Sky House Sky Courtesy: Pushmeet Kohli
Move Making Algorithms Current Solution Search Neighbourhood Optimal Move Energy Solution Space
Higher order CRF models • Higher order models Unary Pairwise Higher-order
Segment-based Potentials No. of pixels not taking l in c
Detector-based Potentials Strength of detector response LuborLadicky, Paul Sturgess, KarteekAlahari, Chris Russell, Philip H.S. Torr, What,Where & How Many? Combining Object Detectors and CRFs, ECCV 2010
Co-occurrence Potentials Global image label set LuborLadicky, Chris Russell, PushmeetKohli, Philip H.S. Torr, Graph Cut based Inference with Co-occurrence Statistics, ECCV, 2010
Joint Stereo + Segmentation Object only potentials Joint potentials LuborLadicky, Paul Sturgess, Chris Russell, SunandoSengupta, Philip H.S. Torr, Joint Optimisation for Object Class Segmentation and Dense Stereo Reconstruction, BMVC 2010
Hierarchical Energies Energy between levels 1 and 0 LuborLadicky, Chris Russell, PushmeetKohli, Philip H.S. Torr, Associative Hierarchical CRFs for Object Class Image Segmentation, ICCV, 2009.
Style-based Potentials Style 1: Style 2: Style-based unary potential Jonathan Warrell, Simon Prince, Philip H.S. Torr, StyP-Boost: A Bilinear Boosting Algorithm for Learning Style-Parameterized Classifiers, BMVC, 2010
TextonBoost (Shotton et al ’09) • Image first convolved with 17-d filter bank • Vectors are clustered, and assigned to ~150 texton indices
TextonBoost (Shotton et al ’09) • Texture-layout features derived from textons • Boosted classifier predicts semantic class
DenseBoost (Ladicky et al ’09) • DenseBoost extends TextonBoost to include • HOG • ColourHOG • Structure / Motion features • State of the art performance on • MSRC (Ladicky et al ’09) • CamVid (Sturgess et al ’09) LuborLadicky, Chris Russell, PushmeetKohli, Philip H.S. Torr, Associative Hierarchical CRFs for Object Class Image Segmentation, ICCV, 2009. Paul Sturgess, KarteekAlahari, LuborLadicky, Philip H.S. Torr, Combining Appearance and Structure from Motion Features for Road Scene Understanding, BMVC, 2009
StyP-Boost Framework (Training) • Training Set • Objective • Classifier form Local features Style Parameters Target vectors
StyP-Boost Framework (Training) • Training Set • Objective • Classifier form Strong learner for class k Loss for class k
StyP-Boost Framework (Training) • Training Set • Objective • Classifier form Weak learner m Style s
Corel: Styles through clustering • Styles found in Corel through clustering 2-styles (98%) 3-styles(96%) 4-styles (89%)
Corel: Styles through clustering • Cluster images based on label histograms during training (2-4 clusters) • Train classifier to predict cluster from image • Use smoothed classifier posteriors as style parameters (training and testing) label label label cluster
Corel: Qualitative results • StyP-Boost reduces noise from classes which don’t co-occur
Corel: Qualitative results • StyP-Boost provides better discrimination of • co-occuring classes
Corel: Quantitative results Test set Training set
Open questions / Problems • Learning from sparsely labeled data Lamp-post Sign
Open questions / Problems • Incorporating 3D and Video Volumetric CRF Ground-plan CRF Image CRF
Open questions / Problems • Using temporal information • Extend detector potentials to include tracking • Use global scene variables for times of day, seasons etc.
Further Questions • Further Questions?