Dressed Human Model DetectionOverviewWhy Would We Want To?ChallengesAppearance VariationsOcclusionModel ClassificationModel Classification (cont.)Slide 9Slide 10Slide 11Slide 12Slide 13Model Classification SummarySlide 15Human Body ModelHuman Body Model (cont.)Slide 18Slide 19Slide 20Slide 21Slide 22Slide 23Slide 24Slide 25Slide 26Slide 27Slide 28Slide 29Slide 30Human Body Model SummarySimilarity MeasureSimilarity Measure (cont.)Slide 34Slide 35Slide 36Bayes’ RuleBayes’ Rule (cont.)Slide 39Bayes’ Rule ExampleBayes’ Rule Example (con’t)Slide 42MAP ~ Maximum A PosterioriSlide 44Slide 45Slide 46The Goodness FunctionThe Goodness Function (cont.)Slide 49Slide 50Slide 51The Goodness Function SummaryDynamic Model Assembling: Step 1Dynamic Model Assembling : Step 2Dynamic Model Assembling : Steps 3&4Body Part IdentificationBody Part Identification (cont.)Multiple Person ExampleMultiple Person Example (cont.)Applying BSM to Human DetectionHuman Detection (cont.)Slide 62Human Detection ResultsHuman Detection Results (cont.)RCR AlgorithmRCR Algorithm (cont.)Slide 67Slide 68Slide 69Slide 70Slide 71Slide 72Slide 73Slide 74Slide 75Slide 76Slide 77RCR Algorithm ExampleRCR Algorithm Example (cont.)ResourcesDressed Human Dressed Human Model DetectionModel DetectionDressed Human Dressed Human Model DetectionModel DetectionBy:Adam SchwarzCAP 5937OverviewOverviewOverviewOverview•Purpose:–Why we want to–What challenges we have to overcome•Means:–Model Classification–Bayesian Similarity Measure (BSM) and Body Part Identification–Recursive Context Reasoning (RCR)Why Would We Want To?Why Would We Want To?Why Would We Want To?Why Would We Want To?•Mobile Robot Navigation–Working safely among humans•Visual Surveillance•Human Motion Capture–Animation, VR, and HCI applications•Shape-Based Image RetrievalChallengesChallengesChallengesChallenges•Appearance Variations–Clothing–Articulation•Occlusion–Other People (crowded areas)–Objects (Partially visible person)•Projection Ambiguities–Projection from 3D model to Image PlaneAppearance VariationsAppearance VariationsAppearance VariationsAppearance VariationsVariations due to ClothingVariations due to ArticulationOcclusionOcclusionOcclusionOcclusionBy an Object By other PeopleModel ClassificationModel ClassificationModel ClassificationModel Classification•Requirements for “good” Object Classification–Independent of Scale, Orientation, and Position–Handle partial occlusions–Allow for articulated moving parts–Handle shape distortions due to noise–Allow for some shape variations–Support efficient shape recognition and classificationModel Classification Model Classification (cont.)(cont.)Model Classification Model Classification (cont.)(cont.)1st: Find Contour OutlinesModel Classification Model Classification (cont.)(cont.)Model Classification Model Classification (cont.)(cont.)2nd: Shape DecompositionRandom NCM NaturalNCM - Negative Curvature Minima (small circles) A good start, but not exactly what we need. So we introduce another constraint…Model Classification Model Classification (cont.)(cont.)Model Classification Model Classification (cont.)(cont.)•Short-Cut Rule: for a valid cut–Must be a straight line–Cross an axis of local symmetry*–Join 2 points on the outline (at least 1 NCM)–Be the shortest cut, if there are several possible competing cuts* Extraneous calculations… alternate method nextModel Classification Model Classification (cont.)(cont.)Model Classification Model Classification (cont.)(cont.)•Salience Constraint replaces axis of symmetry calculation•Salience defines how pronounced a part is, or “part-like” it is•3 Factors determine Salience–Size relative to whole object–Degree to which the part protrudes–Strength of its boundariesModel Classification Model Classification (cont.)(cont.)Model Classification Model Classification (cont.)(cont.)•Let’s take a look at an example and break down the formula:• Cl and Cr have equal arc length based on cut P:PM• Tp represents the threshold determining if a cutmakes a significant part or not• P’ represents a test point for the cut’s endThe Smallest CutSalience of Part( Curve / Cut )Model Classification Model Classification (cont.)(cont.)Model Classification Model Classification (cont.)(cont.)•Final Step: Grouping over-segmented parts–We consider all NCM over a certain magnitude threshold to avoid noise–This may leave extraneous cuts for the same parts–So we group starting with the largest parts first when several possible merges exist–Continue until no more non-decomposable larger parts can be created and all existing large parts can’t be decomposed into significant partsModel Classification Model Classification SummarySummaryModel Classification Model Classification SummarySummaryFind Human’s Silhouette-Contour OutlineSmooth Outline and Find all significant NCM over a magnitude threshold to cut out noiseModel Classification Model Classification SummarySummaryModel Classification Model Classification SummarySummaryUse Short-Cut Method and Salience Constraint to create cutsGroup All Over-Segmented Parts to get Natural Shape DecompositionHuman Body ModelHuman Body ModelHuman Body ModelHuman Body ModelFront ViewSide View•We are working with Images, so 2D models are preferred•We also find that these two models are sufficient•The models also have probability distributions of the spatial relationships between the parts and the torsoHuman Body Model Human Body Model (cont.)(cont.)Human Body Model Human Body Model (cont.)(cont.)•Each body part is modeled by a ribbonWidth (w): the average width along the ribbonLength (l): the major axis from the ribbon spineAspect Ratio (a): w/l, invariant under similarity transforms, captures the global shape while ignoring small local shape deformationsHuman Body Model Human Body Model (cont.)(cont.)Human Body Model Human Body Model (cont.)(cont.)•Aspect Ratio alone is too ambiguous to distinguish different parts (ex. Head and Torso)•The Origin of each part* is located at the joint connecting the part to its parent in the “connect to” hierarchy (more on this in a bit…)*Exception to this is the torso, whose center is its geometric centerHuman Body Model Human Body Model (cont.)(cont.)Human Body Model Human Body Model (cont.)(cont.)General “Connect To” Hierarchy: used as a guide map