Differentiating HOG

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

Wei-Chen Chiu and Mario Fritz, "See the Difference: Direct Pre-Image Reconstruction and Pose Estimation by Differentiating HOG," in IEEE Int’l Conf. on Computer Vision (ICCV), Santiago, Chile, December 13-16, 2015.

@inproceedings{walon2015iccv,
 Author = {Chiu, W.C. and Fritz, M.},
 Title = {See the Difference: Direct Pre-Image Reconstruction and Pose Estimation by Differentiating HOG},
 Booktitle = {IEEE International Conference on Computer Vision (ICCV)},
 Year = {2015},
 Location = {Santiago, Chile}
 }

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 ∇HOG based on the auto-differentiation toolbox Chumpy [1] and show applications to pre-image visualization and pose estimation which extends the existing differentiable renderer OpenDR [2] pipeline. Both applications improve on the respective state-of-the-art HOG approaches.

We exploit the piecewise differentiability of the popular HOG descriptor for end-to-end optimization. The figure shows applications on the pre-image reconstruction given HOG features as well as the pose estimation task based on the same idea.

Material

Reference

[1] M. Loper. Chumpy. https://github.com/ mattloper/chumpy.

[2] M. M. Loper and M. J. Black. Opendr: An approximate differentiable renderer. In ECCV. 2014.