Monocular Segment-Wise Depth: Monocular Depth Estimation Based on a Semantic Segmentation Prior

Atapour-Abarghouei, A.; Breckon, T.P.

Abstract

Monocular depth estimation using novel learning-based approaches has recently emerged as a promising potential alternative to more conventional 3D scene capture technologies within real-world scenarios. Many such solutions often depend on large quantities of ground truth depth data, which is rare and often intractable to obtain. Others attempt to estimate disparity as an intermediary step using a secondary supervisory signal, leading to blurring and other undesirable artefacts. In this paper, we propose a monocular depth estimation approach, which employs a jointly-trained pixel-wise semantic understanding step to estimate depth for individuallyselected groups of objects (segments) within the scene. The separate depth outputs are efficiently fused to generate the final result. This creates more simplistic learning objectives for the jointly-trained individual networks, leading to more accurate overall depth. Extensive experimentation demonstrates the efficacy of the proposed approach compared to contemporary state-of-the-art techniques within the literature.

Citation

Atapour-Abarghouei, A., & Breckon, T. (2019, September). Monocular Segment-Wise Depth: Monocular Depth Estimation Based on a Semantic Segmentation Prior. Presented at IEEE International Conference on Image Processing, Taipei, Taiwen

• Presentation Conference Type: Conference Paper (published)
• Conference Name: IEEE International Conference on Image Processing
• Start Date: Sep 22, 2019
• End Date: Sep 25, 2019
• Acceptance Date: Apr 30, 2019
• Publication Date: Sep 1, 2019
• Deposit Date: Jun 4, 2019
• Publicly Available Date: Nov 12, 2019
• Pages: 4295-4299
• Series ISSN: 2381-8549
• Book Title: 2019 IEEE International Conference on Image Processing (ICIP) ; proceedings.
• DOI: https://doi.org/10.1109/icip.2019.8803551
• Public URL: https://durham-repository.worktribe.com/output/1144133
• Additional Information: Conference dates September 22-25 2019. Not sure about pub date. IEEE xplore states: Date Added to IEEE Xplore: 26 August 2019 - but that's before the conference. [