A 2D Human Body Model Dressed in Eigen Clothing

Abstract

Abstract. Detection, tracking, segmentation and pose estimation of people in monocular images are widely studied. Two-dimensional models of the human body are extensively used, however, they are typically fairly crude, representing the body either as a rough outline or in terms of articulated geometric primitives. We describe a new 2D model of the human body contour that combines an underlying naked body with a lowdimensional clothing model. The naked body is represented as a Contour Person that can take on a wide variety of poses and body shapes. Clothing is represented as a deformation from the underlying body contour. This deformation is learned from training examples using principal component analysis to produce eigen clothing. We find that the statistics of clothing deformations are skewed and we model the a priori probability of these deformations using a Beta distribution. The resulting generative model captures realistic human forms in monocular images and is used to infer 2D body shape and pose under clothing. We also use the coefficients of the eigen clothing to recognize different categories of clothing on dressed people. The method is evaluated quantitatively on synthetic and real images and achieves better accuracy than previous methods for estimating body shape under clothing.

Learning people detection models from few training samples

“…In fact, using recordings from several hundreds of people should allow to reach performance levels that are beyond what can be reached with today's manually and tediously labeled data. To further increase the variability in appearance we also envision the combination of the Reshape data generation with an additional model for clothing generation such as Eigen Clothing [19].…”

Parametrized shape models for clothing

“…Few methods investigate clothing explicitly. Notable exceptions to this is the recent work of [8], which expands the SCAPE manifold-based model [1] which attempt to account for clothing variation [18], and methods for estimating folds in deformable surfaces [22]. In this work we adopt a much simpler model, and propose schemes for direct optimization of specific shape forms that can be directly related to ensuing manipulation; we expect our method to become even more robust when the above statistical shape and observation model methods are applied to our approach, but the results with classic methods prove to be significant as shown below.…”

Retrieving Similar Styles to Parse Clothing

“…Early work on clothing representation modeled clothing as a grammar of sketch templates [30]. Other work took a subspace approach to describe clothing deformations [31], [32], or deformable spatial priors [33]. These approaches focus mainly on how to model shape deformations for clothing recognition.…”