See the Difference: Direct Pre-Image Reconstruction and Pose Estimation by Differentiating HOG

TL;DR

This paper introduces a differentiable HOG descriptor enabling end-to-end optimization for pre-image visualization and pose estimation, leading to improved results over existing methods.

Contribution

It presents a differentiable implementation of HOG using auto-differentiation, facilitating advanced analysis and optimization in vision tasks.

Findings

Enhanced pre-image visualization accuracy

Improved pose estimation performance

Integration with differentiable rendering pipeline

Abstract

The Histogram of Oriented Gradient (HOG) descriptor has led to many advances in computer vision over the last decade and is still part of many state of the art approaches. We realize that the associated feature computation is piecewise differentiable and therefore many pipelines which build on HOG can be made differentiable. This lends to advanced introspection as well as opportunities for end-to-end optimization. We present our implementation of $\nabla$HOG based on the auto-differentiation toolbox Chumpy and show applications to pre-image visualization and pose estimation which extends the existing differentiable renderer OpenDR pipeline. Both applications improve on the respective state-of-the-art HOG approaches.