Due to the significant non-linear distortions of fisheye camera models, supervised learning techniques for monocular depth estimation usually do not generalize well when applied to this kind of projection models. Additionally, there is no publicly available dataset for monocular fisheye camera depth estimation emphasizing automated valet parking. As a result, we initiate a unique depth estimate fisheye dataset optimized for automated valet parking. On this premise, we also construct a system for monocular fisheye depth estimation using an enhanced deep optimizer for improving the results of supervised monocular depth estimation on fisheye camera images. The proposed method produces impressive results on our ZMDepth dataset. We also evaluate our optimizer-based method on the KITTI dataset and achieved state-of-the-art results. We will make a partial ZMDepth dataset publicly available at: https://github.com/NamespaceMain/ZMDepth.
Monocular Fisheye Depth Estimation for Automated Valet Parking: Dataset, Baseline and Deep Optimizers
08.10.2022
2524558 byte
Aufsatz (Konferenz)
Elektronische Ressource
Englisch
AUTOMATED VALET PARKING METHOD AND AUTOMATED VALET PARKING SYSTEM
Europäisches Patentamt | 2022
|AUTOMATED VALET PARKING METHOD AND AUTOMATED VALET PARKING SYSTEM
Europäisches Patentamt | 2024
|VEHICLE, AUTOMATED VALET PARKING SYSTEM, AND AUTOMATED VALET PARKING METHOD
Europäisches Patentamt | 2023
|Autonomous valet parking system and method using fisheye lens camera
Europäisches Patentamt | 2023