SECOND: Sparsely Embedded Convolutional Detection
By Yan Yan, Yuxing Mao, Bo Li. 2018
3,575 citations recorded by the source corpus.
LiDAR-based or RGB-D-based object detection is used in numerous applications, ranging from autonomous driving to robot vision. Voxel-based 3D convolutional networks have been used for some time to enhance the retention of information when processing point cloud LiDAR data. However, problems remain, including a slow inference speed and low orientation estimation performance. We therefore investigate an improved sparse convolution method for such networks, which significantly increases the speed of both training and inference. We also introduce a new form of angle loss regression to improve the orientation estimation performance and a new data augmentation approach that can enhance the convergence speed and performance. The proposed network produces state-of-the-art results on the KITTI 3D object detection benchmarks while maintaining a fast inference speed.
Subfields: Aerospace Engineering, Computer Vision and Pattern Recognition, Environmental Engineering
How researchers see this paper
The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.
In Aerospace Engineering
- Philipp Krahenbuhl
- Hongsheng Li
- Shaoshuai Shi
- Jiaya Jia
- Jian‐Yuan Sun
- Xiaogang Wang
- Xingyi Zhou
- Guan Huang
In Computer Vision and Pattern Recognition
- Luc Van Gool
- Xiaogang Wang
- Dacheng Tao
- Ming-Hsuan Yang
- Trevor Darrell
- Alan L. Yuille
- Jitendra Malik
- Bernt Schiele
In Environmental Engineering
- Shaoshuai Shi
- Qi Chen
- Philipp Krahenbuhl
- Xingyi Zhou
- Tianwei Yin
- Xiaogang Wang
- Hongsheng Li
- Li Jiang
Citations in the corpus
- Static-Dynamic Co-Teaching for Class-Incremental 3D Object Detection (2021, background): “Many deep learning-based approaches (Li, Zhang, and Xia 2016; Chen et al. 2017; Beltrán et al. 2018; Yan, Mao, and Li 2018; Yang, Luo, and Urtasun 2018; Zeng et al. 2018; Zhou and Tuzel 2018; Chen et al. 2019; Lang et al. 2019; Qi et al. 2019; Shi, Wang, and Li 2019; Yang et al. 2019; Zhou et al.…”
- A Survey of Autonomous Driving: Common Practices and Emerging Technologies (2019, methodology): “2D methods are far less computationally expensive, but recent methods that take point sparsity into account [126] are real-time viable and rapidly”
- ST3D: Self-training for Unsupervised Domain Adaptation on 3D Object Detection (2021, methodology): “Performance of ST3D on Waymo → KITTI task using SECOND-IoU [46], compared to other unsupervised (i.”
- DeepFusion: Lidar-Camera Deep Fusion for Multi-Modal 3D Object Detection (2022, methodology): “SECOND [40] builds on VoxelNet and proposes using sparse 3D convolutions to increase efficiency.”
- TransFusion: Robust LiDAR-Camera Fusion for 3D Object Detection with Transformers (2022, methodology): “Note that we also find the copy-andpaste augmentation strategy [52] benefits the convergence”
- BEVFormer: Learning Bird's-Eye-View Representation from Multi-Camera Images via Spatiotemporal Transformers (2022, methodology): “Despite the remarkable progress of LiDAR-based methods [43, 20, 53, 49, 8], camera-based approaches [45, 32, 47, 30] have attracted extensive attention in recent years.”
- Class-balanced Grouping and Sampling for Point Cloud 3D Object Detection (2019, methodology): “Apart from regular classification and bounding box regression branch required by 3D object detection, we add an orientation classification branch as proposed in SECOND [28].”
- STD: Sparse-to-Dense 3D Object Detector for Point Cloud (2019, methodology): “97% compared to PointRCNN [30], PointPillars [15], and SECOND [34] respectively on the moderate set.”
- SE-SSD: Self-Ensembling Single-Stage Object Detector From Point Cloud (2021, methodology): “SECOND [30] modifies the sparse convolution [7, 15] to efficiently extract features from sparse voxels.”
- TANet: Robust 3D Object Detection from Point Clouds with Triple Attention (2019, methodology): “Ground-truth boxes are randomly sampled and placed into raw samples to simulate the scenes crowded with multiple objects (Yan, Mao, and Li 2018).”
- Unifying Voxel-based Representation with Transformer for 3D Object Detection (2022, background): “Final predictions are usually generated on top of the bird-eye view (BEV) space with the flatted height axis [22, 23, 24].”
- CenterFusion: Center-based Radar and Camera Fusion for 3D Object Detection (2020, methodology): “The majority of LiDAR-based methods either use 3D voxels [12, 35] or 2D projections [13, 5, 29, 31] for point cloud representation.”
- F-cooper: feature based cooperative perception for autonomous vehicle edge computing system using 3D point clouds (2019, background): “Based on our best acknowledge, the state of the art 3D object detection precision based on monocular LiDAR (Light Detection and Ranging) data comes from VoxelNet [37], SECOND [36] and PointR-CNN [31], etc.”
- Center-based 3D Object Detection and Tracking (2020, methodology): “Modern 3D object detectors [19, 27, 54, 64] uses a 3D encoder that quantizes the point-cloud into regular bins.”
- CIA-SSD: Confident IoU-Aware Single-Stage Object Detector From Point Cloud (2020, methodology): “Following previous works, e.g. , SASSD (He et al. 2020) and SECOND (Yan, Mao, and Li 2018), we conducted experiments on the most commonly-used car category and evaluated the results by average precision (AP) with IoU threshold 0.7.”
- Point-Voxel CNN for Efficient 3D Deep Learning (2019, background): “There are PointPillars [21] and SECOND [48] focusing on efficient 3D detection.”