Difficulty-Aware Label-Guided Denoising for Monocular 3D Object Detection
This addresses a fundamental limitation in cost-effective 3D perception for autonomous driving and robotics, though it appears incremental as it builds on existing DETR-based methods.
The paper tackles inaccurate depth estimation in monocular 3D object detection by proposing MonoDLGD, a framework that adaptively perturbs and reconstructs ground-truth labels based on detection difficulty, achieving state-of-the-art performance on the KITTI benchmark across all difficulty levels.
Monocular 3D object detection is a cost-effective solution for applications like autonomous driving and robotics, but remains fundamentally ill-posed due to inherently ambiguous depth cues. Recent DETR-based methods attempt to mitigate this through global attention and auxiliary depth prediction, yet they still struggle with inaccurate depth estimates. Moreover, these methods often overlook instance-level detection difficulty, such as occlusion, distance, and truncation, leading to suboptimal detection performance. We propose MonoDLGD, a novel Difficulty-Aware Label-Guided Denoising framework that adaptively perturbs and reconstructs ground-truth labels based on detection uncertainty. Specifically, MonoDLGD applies stronger perturbations to easier instances and weaker ones into harder cases, and then reconstructs them to effectively provide explicit geometric supervision. By jointly optimizing label reconstruction and 3D object detection, MonoDLGD encourages geometry-aware representation learning and improves robustness to varying levels of object complexity. Extensive experiments on the KITTI benchmark demonstrate that MonoDLGD achieves state-of-the-art performance across all difficulty levels.