The KITTI Vision Benchmark Suite

Currently, in Autonomous Driving (AD), most of
the 3D object detection frameworks (either
anchor- or anchor-freebased) consider the
detection as a Bounding Box (BBox) regression
problem. However, this compact representation is
not sufficient to explore all the information of
the objects. To tackle this problem, we propose a
simple but practical detection framework to
jointly predict the 3D BBox and instance
segmentation. For instance segmentation, we
propose a Spatial Embeddings (SEs) strategy to
assemble all foreground points into their
corresponding object centers. Base on the SE
results, the object proposals can be generated
based on a simple clustering strategy. For each
cluster, only one proposal is generated.
Therefore, the NonMaximum Suppression (NMS)
process is no longer needed here. Finally, with
our proposed instance-aware ROI pooling, the BBox
is refined by a second-stage network.

tbd

@inproceedings{zhou2020joint,
title={Joint 3D Instance Segmentation and
Object Detection for Autonomous Driving},
author={Zhou, Dingfu and Fang, Jin and Song,
Xibin and Liu, Liu and Yin, Junbo and Dai, Yuchao
and Li, Hongdong and Yang, Ruigang},
booktitle={Proceedings of the IEEE/CVF
Conference on Computer Vision and Pattern
Recognition},
pages={1839--1849},
year={2020}
}