Advancements in deep learning-based 3D object detection necessitate the availability of large-scale datasets. However, this requirement introduces the challenge of manual annotation, which is often both burdensome and time-consuming. To tackle this issue, the literature has seen the emergence of several pseudolabelingbased frameworks for 3D object detection which can automatically annotate unlabeled data. Nevertheless, these generated pseudo labels may contain varying degree of noise and may not be as accurate as those labeled by humans. In this thesis, we present the first approach that addresses the inherent ambiguities present in 3D box pseudo labels by introducing an Evidential Deep Learning (EDL) based uncertainty estimation framework. Specifically, we propose MEDLU, an EDL framewo...[ Read more ]

Advancements in deep learning-based 3D object detection necessitate the availability of large-scale datasets. However, this requirement introduces the challenge of manual annotation, which is often both burdensome and time-consuming. To tackle this issue, the literature has seen the emergence of several pseudolabelingbased frameworks for 3D object detection which can automatically annotate unlabeled data. Nevertheless, these generated pseudo labels may contain varying degree of noise and may not be as accurate as those labeled by humans. In this thesis, we present the first approach that addresses the inherent ambiguities present in 3D box pseudo labels by introducing an Evidential Deep Learning (EDL) based uncertainty estimation framework. Specifically, we propose MEDLU, an EDL framework based on MTrans, which not only generates pseudo labels but also quantifies the associated uncertainties. We observe that applying EDL to 3D object detection presents three primary challenges: (1) degraded pseudo label quality compared to other autolabelers; (2) excessively high evidential uncertainty estimates; and (3) lack of clear interpretability and effective utilization of uncertainties for downstream tasks. We tackle these issues through the introduction of an uncertainty-aware IoU-based loss, an evidence-aware multi-task loss, and the implementation of a post-processing stage for uncertainty refinement. Our experimental results demonstrate that probabilistic detectors trained using the outputs of MEDL-U surpass deterministic detectors trained using outputs from previous 3D annotators on the KITTI val set for all difficulty levels. Moreover, MEDL-U achieves state-of-the-art results on the KITTI official test set compared to existing 3D automatic annotators.