Conditional Random Fields for Object Recognition

TLDR

Object recognition often relies on local features assumed to be conditionally independent, an assumption that may be too restrictive for many object classes. The authors aim to develop a discriminative part‑based approach that extends Conditional Random Fields with hidden variables for unified part‑based object recognition in cluttered scenes. Objects are modeled as flexible constellations of parts conditioned on local observations, with each class’s part assignment probability modeled by a CRF whose parameters are estimated by maximum likelihood and recognition achieved by selecting the most likely class. The proposed CRF framework relaxes the conditional independence assumption of local features, enhancing flexibility for a wide range of object classes.

Abstract

We present a discriminative part-based approach for the recognition of object classes from unsegmented cluttered scenes. Objects are modeled as flexible constellations of parts conditioned on local observations found by an interest operator. For each object class the probability of a given assignment of parts to local features is modeled by a Conditional Random Field (CRF). We propose an extension of the CRF framework that incorporates hidden variables and combines class conditional CRFs into a unified framework for part-based object recognition. The parameters of the CRF are estimated in a maximum likelihood framework and recognition proceeds by finding the most likely class under our model. The main advantage of the proposed CRF framework is that it allows us to relax the assumption of conditional independence of the observed data (i.e. local features) often used in generative approaches, an assumption that might be too restrictive for a considerable number of object classes.