Discovering Useful Parts for Pose Estimation in Sparsely Annotated Datasets

TL;DR

This paper presents a method to improve pose estimation accuracy by discovering and utilizing unannotated parts in training images, enhancing traditional models and demonstrating significant accuracy improvements on a new dataset.

Contribution

Introduces a novel approach to discover unannotated parts for pose estimation, improving accuracy and applicability over existing methods.

Findings

Landmark localization at least twice as accurate as baseline

Significant improvement over previous methods on hawkmoth images

Proposed approach is more generally applicable

Abstract

Our work introduces a novel way to increase pose estimation accuracy by discovering parts from unannotated regions of training images. Discovered parts are used to generate more accurate appearance likelihoods for traditional part-based models like Pictorial Structures [13] and its derivatives. Our experiments on images of a hawkmoth in flight show that our proposed approach significantly improves over existing work [27] for this application, while also being more generally applicable. Our proposed approach localizes landmarks at least twice as accurately as a baseline based on a Mixture of Pictorial Structures (MPS) model. Our unique High-Resolution Moth Flight (HRMF) dataset is made publicly available with annotations.