Exploring intermediate representation for monocular vehicle pose estimation

TL;DR

This paper introduces a novel deep learning framework that uses intermediate geometrical representations and a geometry-aware loss to improve monocular vehicle pose estimation from a single RGB image, outperforming previous methods.

Contribution

It proposes a new approach utilizing intermediate geometrical representations and a projective invariant loss to enhance vehicle pose estimation accuracy.

Findings

Outperforms previous monocular methods on KITTI benchmark

Achieves performance comparable to stereo methods

Utilizes unlabeled data effectively during training

Abstract

We present a new learning-based framework to recover vehicle pose in SO(3) from a single RGB image. In contrast to previous works that map from local appearance to observation angles, we explore a progressive approach by extracting meaningful Intermediate Geometrical Representations (IGRs) to estimate egocentric vehicle orientation. This approach features a deep model that transforms perceived intensities to IGRs, which are mapped to a 3D representation encoding object orientation in the camera coordinate system. Core problems are what IGRs to use and how to learn them more effectively. We answer the former question by designing IGRs based on an interpolated cuboid that derives from primitive 3D annotation readily. The latter question motivates us to incorporate geometry knowledge with a new loss function based on a projective invariant. This loss function allows unlabeled data to be used in the training stage to improve representation learning. Without additional labels, our system outperforms previous monocular RGB-based methods for joint vehicle detection and pose estimation on the KITTI benchmark, achieving performance even comparable to stereo methods. Code and pre-trained models are available at this https URL.