Probabilistic Structure

1. Level I Level III Level II SMC Proposal map Qk is computed from line segments in the lower part of the body that have not been visited by SMC (ek+1:K). Prior (proposal) k and likelihood (weight) k are computed from line segments in the upper part that have already been visited (e1:k-1). DP Calf Face Scale = 2 Scale = 2 Scale = 3 Scale = 3 Scale = 4 Scale = 4 Body Localization in Still Images Using Hierarchical Models and Hybrid Search Jiayong Zhang1, Jiebo Luo2, Robert Collins3, Yanxi Liu1 1Carnegie Mellon University 2Kodak Research Lab3Pennsylvania State University • Highlights • Generic setting: single image, arbitrary pose & viewpoint • 3-level model decomposition (2D, landmark-based) • Combination of deterministic and stochastic search • Assurance of both robustness and accuracy • Assumptions • Torso ║ imaging plane  No external occlusion Experiments  Trained on CMU Mobo  Tested on 340 images, success rate ≈ 40%  Complete candidate sets clustered into hyper-modes (sorted by likelihood scores). The ideal automated scoring function would have the top-ranked mode (leftmost in each row) coinciding with the preferred mode (with yellow frame).  Hierarchical Model Decomposition (I) View-independent tree-structured model. (II) Mixture model, with eight view-dependent components from angles uniformly distributed in [0,2]. (III) Expanded mixture model with increased landmark density.  Hybrid Strategy Combining DP & SMC Step 1. Fit (I) by Dynamic Programming (DP). The search starts from the bottom (feet) to the top (head). The output is a series of proposal maps, together with foreground masks for different body parts. Step 2. Fit (II) by Sequential Monte Carlo (SMC). The search starts from the top (head) to the bottom (feet). Proposal maps from DP are combined with the prior terms of the mixture model into an improved proposal function for the SMC search, while the foreground masks generated from DP are utilized in the computation of SMC importance weights. Step 3. Fit (III) by local optimization, initialized by the SMC output. Combining Top-down & Bottom-up Bottom-up Detection of Partial Bodies  Physical Topology  Examples assembled from the “preferred” modes.  Probabilistic Structure Marginal Posteriors: (Proposal Maps) Visualize {Q} Foreground masks from {Q}