RangeRCNN: Towards Fast and Accurate 3D Object Detection with Range Image Representation

Abstract

We present RangeRCNN, a novel and effective 3D object detection framework based on the range image representation. Most existing 3D object detection methods are either voxel-based or point-based. Though several optimizations have been introduced to ease the sparsity issue and speed up the running time, the two representations are still computationally inefficient. Compared to these two representations, the range image representation is dense and compact which can exploit the powerful 2D convolution and avoid the uncertain receptive field caused by the sparsity issue. Even so, the range image representation is not preferred in 3D object detection due to the scale variation and occlusion. In this paper, we utilize the dilated residual block to better adapt different object scales and obtain a more flexible receptive field on range image. Considering the scale variation and occlusion of the range image, we propose the RV-PV-BEV (Range View to Point View to Bird's Eye View) module to transfer the feature from the range view to the bird's eye view. The anchor is defined in the BEV space which avoids the scale variation and occlusion. Both RV and BEV cannot provide enough information for height estimation, so we propose a two-stage RCNN for better 3D detection performance. The point view aforementioned does not only serve as a bridge from RV to BEV but also provides pointwise features for RCNN. Extensive experiments show that the proposed RangeRCNN achieves state-of-the-art performance on the KITTI 3D object detection dataset. We prove that the range image based methods can be effective on the KITTI dataset which provides more possibilities for real-time 3D object detection.

Survey and Systematization of 3D Object Detection Models and Methods

“…Since the LiDAR projection is still neither in a processable state nor contains any discriminative features such as RGB information, the projected RV is partitioned into a finegrained grid and encoded in the successive feature initialization step (see Section 7.4.1). Meyer et al (2019b) and Liang et al (2020) emphasize that the naturally compact RV results in a more efficient computation in comparison to other projections. In addition, the information loss of projection is considerably small since the RV constitutes the native representation of a rotating LiDAR sensor (Liang et al, 2020).…”

“…Meyer et al (2019b) and Liang et al (2020) emphasize that the naturally compact RV results in a more efficient computation in comparison to other projections. In addition, the information loss of projection is considerably small since the RV constitutes the native representation of a rotating LiDAR sensor (Liang et al, 2020). At the same time, RV suffers from distorted object size and shape on account of its cylindrical image character (Yang et al, 2018b).…”

“…Exemplary models using BEV representation can be found in the work from Wang et al (2018), Beltrán et al (2018), Liang et al (2018), Yang et al (2018b), Simon et al (2019b), Zeng et al (2018), , Ali et al (2019), He et al (2020), Zheng et al (2020), Liang et al (2020), and.…”

“…Often BEV and RV are not used as single data representation but as a multi view approach, meaning that RV and BEV but also images of monocular nature are combined to represent the spatial information of a point cloud. Chen et al (2017) were the first to integrate this concept into a 3DOD pipeline, followed by many other models adapting to fuse 2D representations from different perspectives (e.g., Ku et al, 2018;Liang et al, 2020).…”

“…Performancewise, Kuang et al (2020) marily get filled with statistical quantities of the within lying points. Only more recently, first models started to encode the features by deep learning methods (e.g., Lehner et al, 2019;Zheng et al, 2020;Liang et al, 2020).…”