The KITTI Vision Benchmark Suite

First, we train a single-stage convolutional
neural network (CNN) that produces multiple visual
and geometric cues of interest: 2D bounding
boxes, 2D keypoints of interest, coarse object
orientations and object dimensions. Subsets of
these cues are then used to poll probable
ground planes from a pre-computed database of
ground planes, to identify the “best fit” plane
with highest consensus. Once identified, the
“best fit” plane provides enough constraints to
successfully construct the desired 3D detection
box, without directly predicting the 6DoF
pose of the object. The entire ground plane
polling (GPP) procedure is constructed as a non-
parametrized layer of the CNN that outputs the
desired “best fit” plane and the corresponding 3D
keypoints, which together define the final 3D
bounding box.

NA

@article{rangesh2020ground,
title={Ground plane polling for 6dof pose
estimation of objects on the road},
author={Rangesh, Akshay and Trivedi, Mohan M},
journal={IEEE Transactions on Intelligent
Vehicles},
year={2020},
publisher={IEEE}
}