Stereo Frustums: A Siamese Pipeline for 3D Object Detection

TL;DR

This paper introduces a lightweight stereo frustums matching module that leverages 2D proposals and epipolar constraints to improve 3D object detection accuracy for autonomous vehicles, outperforming existing stereo-based methods.

Contribution

It presents a novel Siamese pipeline that directly matches 2D proposals across stereo images using epipolar geometry, bypassing traditional stereo matching for 3D detection.

Findings

Outperforms state-of-the-art stereo 3D detection methods on KITTI dataset

Uses a novel matching algorithm based on epipolar constraints

Achieves accurate 3D bounding boxes with a lightweight framework

Abstract

The paper proposes a light-weighted stereo frustums matching module for 3D objection detection. The proposed framework takes advantage of a high-performance 2D detector and a point cloud segmentation network to regress 3D bounding boxes for autonomous driving vehicles. Instead of performing traditional stereo matching to compute disparities, the module directly takes the 2D proposals from both the left and the right views as input. Based on the epipolar constraints recovered from the well-calibrated stereo cameras, we propose four matching algorithms to search for the best match for each proposal between the stereo image pairs. Each matching pair proposes a segmentation of the scene which is then fed into a 3D bounding box regression network. Results of extensive experiments on KITTI dataset demonstrate that the proposed Siamese pipeline outperforms the state-of-the-art stereo-based 3D bounding box regression methods.