A General Pipeline for 3D Detection of Vehicles
This work addresses 3D perception for autonomous vehicles, but it is incremental as it builds on existing 2D methods with minimal changes.
The paper tackles the problem of 3D vehicle detection for autonomous driving by proposing a flexible pipeline that adapts 2D detection networks to fuse with 3D point clouds, achieving second-best results on the KITTI dataset.
Autonomous driving requires 3D perception of vehicles and other objects in the in environment. Much of the current methods support 2D vehicle detection. This paper proposes a flexible pipeline to adopt any 2D detection network and fuse it with a 3D point cloud to generate 3D information with minimum changes of the 2D detection networks. To identify the 3D box, an effective model fitting algorithm is developed based on generalised car models and score maps. A two-stage convolutional neural network (CNN) is proposed to refine the detected 3D box. This pipeline is tested on the KITTI dataset using two different 2D detection networks. The 3D detection results based on these two networks are similar, demonstrating the flexibility of the proposed pipeline. The results rank second among the 3D detection algorithms, indicating its competencies in 3D detection.