Monocular 3D Object Detection for Autonomous Driving

TLDR

The paper aims to perform monocular 3D object detection for autonomous driving, focusing on generating class‑specific 3D proposals. The approach generates 3D object proposals via energy minimization that enforces ground‑plane placement, then scores each candidate using potentials based on semantic segmentation, context, size/location priors, and typical shape before feeding them to a CNN. Experiments show the proposal generation method outperforms all monocular competitors and achieves the best detection performance on the KITTI benchmark.

Abstract

The goal of this paper is to perform 3D object detection from a single monocular image in the domain of autonomous driving. Our method first aims to generate a set of candidate class-specific object proposals, which are then run through a standard CNN pipeline to obtain high-quality object detections. The focus of this paper is on proposal generation. In particular, we propose an energy minimization approach that places object candidates in 3D using the fact that objects should be on the ground-plane. We then score each candidate box projected to the image plane via several intuitive potentials encoding semantic segmentation, contextual information, size and location priors and typical object shape. Our experimental evaluation demonstrates that our object proposal generation approach significantly outperforms all monocular approaches, and achieves the best detection performance on the challenging KITTI benchmark, among published monocular competitors.