Triangulation Learning Network: from Monocular to Stereo 3D Object Detection

06/04/2019
by   Zengyi Qin, et al.
0

In this paper, we study the problem of 3D object detection from stereo images, in which the key challenge is how to effectively utilize stereo information. Different from previous methods using pixel-level depth maps, we propose employing 3D anchors to explicitly construct object-level correspondences between the regions of interest in stereo images, from which the deep neural network learns to detect and triangulate the targeted object in 3D space. We also introduce a cost-efficient channel reweighting strategy that enhances representational features and weakens noisy signals to facilitate the learning process. All of these are flexibly integrated into a solid baseline detector that uses monocular images. We demonstrate that both the monocular baseline and the stereo triangulation learning network outperform the prior state-of-the-arts in 3D object detection and localization on the challenging KITTI dataset.

READ FULL TEXT

page 3

page 4

page 7

page 8

research
12/03/2021

SGM3D: Stereo Guided Monocular 3D Object Detection

Monocular 3D object detection is a critical yet challenging task for aut...
research
07/05/2023

SVDM: Single-View Diffusion Model for Pseudo-Stereo 3D Object Detection

One of the key problems in 3D object detection is to reduce the accuracy...
research
03/23/2021

Stereo Object Matching Network

This paper presents a stereo object matching method that exploits both 2...
research
03/16/2022

MonoJSG: Joint Semantic and Geometric Cost Volume for Monocular 3D Object Detection

Due to the inherent ill-posed nature of 2D-3D projection, monocular 3D o...
research
04/24/2023

Transformer-based stereo-aware 3D object detection from binocular images

Vision Transformers have shown promising progress in various object dete...
research
04/19/2022

Shape-Aware Monocular 3D Object Detection

The detection of 3D objects through a single perspective camera is a cha...
research
08/25/2020

MonStereo: When Monocular and Stereo Meet at the Tail of 3D Human Localization

Monocular and stereo vision are cost-effective solutions for 3D human lo...

Please sign up or login with your details

Forgot password? Click here to reset