Learning-Deep-Learning

GPP: Ground Plane Polling for 6DoF Pose Estimation of Objects on the Road

November 2019

tl;dr: Regress tireline and height and project to the best ground plane near the car.

Overall impression

GPP generates artificial 2D landmarks with 3D bbox annotation. It purposefully predicts more attributes than needed to estimate 3D bbox (over-determined) and uses these predictions to from maximum consensus set of attributes, in a manner similar to RANSAC, making it more robust to outliers.

Key ideas

Technical details

Notes