Spotlight the Negatives: A Generalized Discriminative Latent Model

TL;DR

This paper introduces a generalized discriminative latent variable model that incorporates negative latent variables for background, enhancing visual recognition performance by modeling both foreground and background variations.

Contribution

It formalizes the GLVM scoring function and demonstrates its benefits, including significant improvements in detection tasks over traditional models.

Findings

Significant performance improvements on detection tasks.

Theoretical benefits for existing visual recognition methods.

Effective modeling of background variations enhances accuracy.

Abstract

Discriminative latent variable models (LVM) are frequently applied to various visual recognition tasks. In these systems the latent (hidden) variables provide a formalism for modeling structured variation of visual features. Conventionally, latent variables are de- fined on the variation of the foreground (positive) class. In this work we augment LVMs to include negative latent variables corresponding to the background class. We formalize the scoring function of such a generalized LVM (GLVM). Then we discuss a framework for learning a model based on the GLVM scoring function. We theoretically showcase how some of the current visual recognition methods can benefit from this generalization. Finally, we experiment on a generalized form of Deformable Part Models with negative latent variables and show significant improvements on two different detection tasks.