Reinforced Axial Refinement Network for Monocular 3D Object Detection
This work addresses the challenge of 3D object detection from 2D images for applications like autonomous driving, but it is incremental as it builds on existing methods as a post-processing stage.
The paper tackles the problem of monocular 3D object detection by proposing a reinforcement learning-based refinement method to improve sampling efficiency, achieving performance gains on the KITTI dataset with minimal computational overhead.
Monocular 3D object detection aims to extract the 3D position and properties of objects from a 2D input image. This is an ill-posed problem with a major difficulty lying in the information loss by depth-agnostic cameras. Conventional approaches sample 3D bounding boxes from the space and infer the relationship between the target object and each of them, however, the probability of effective samples is relatively small in the 3D space. To improve the efficiency of sampling, we propose to start with an initial prediction and refine it gradually towards the ground truth, with only one 3d parameter changed in each step. This requires designing a policy which gets a reward after several steps, and thus we adopt reinforcement learning to optimize it. The proposed framework, Reinforced Axial Refinement Network (RAR-Net), serves as a post-processing stage which can be freely integrated into existing monocular 3D detection methods, and improve the performance on the KITTI dataset with small extra computational costs.