We introduce a generic and efficient method for 2D and 3D shape estimation via density fields. Our method models shape as a density map and uses the notion of density to fit a model to a rapidly computed occupancy map of the foreground object.
We show how to utilize hierarchical (pyramid-like) object segmentation data to regularize a hierarchical model fitting. With primary focus on estimating 3D shapes of non-rigid articulated objects such as human bodies, we illustrate our approach with examples of efficient model fitting to 3D occupancy maps of human figures. We also discuss a number of extensions of our method to applications involving non-rigid object tracking and movement analysis.
Recently we have started developing a radial-based density framework for learning and fitting models to shapes of various complexities. This framework allows to model shapes in a systematic way and offers numerous performance benefits with respect to our previous approach. In addition, we are exploring density models being fitted to grayscale volume and area data.
Major contributions
- generic density-based shape modeling framework
- hierarchical model fitting
- shape modeling to any precision
Differences from similar approaches
- models with low number of degrees of freedom
- works with volumetric shape data
- generalizes to any number of dimensions
Applications
- recognition and tracking of non-rigid articulated objects
- physically realistic modeling and visualization
- flexible and efficient shape representation
real-world data
corresponding models