Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Miangoleh, S. Mahdi H., Reddy, Mahesh, Aksoy, Yağız
Format: Preprint
Veröffentlicht: 2024
Schlagworte:
Online-Zugang:https://arxiv.org/abs/2406.09374
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
_version_ 1866913390628175872
author Miangoleh, S. Mahdi H.
Reddy, Mahesh
Aksoy, Yağız
author_facet Miangoleh, S. Mahdi H.
Reddy, Mahesh
Aksoy, Yağız
contents Existing methods for scale-invariant monocular depth estimation (SI MDE) often struggle due to the complexity of the task, and limited and non-diverse datasets, hindering generalizability in real-world scenarios. This is while shift-and-scale-invariant (SSI) depth estimation, simplifying the task and enabling training with abundant stereo datasets achieves high performance. We present a novel approach that leverages SSI inputs to enhance SI depth estimation, streamlining the network's role and facilitating in-the-wild generalization for SI depth estimation while only using a synthetic dataset for training. Emphasizing the generation of high-resolution details, we introduce a novel sparse ordinal loss that substantially improves detail generation in SSI MDE, addressing critical limitations in existing approaches. Through in-the-wild qualitative examples and zero-shot evaluation we substantiate the practical utility of our approach in computational photography applications, showcasing its ability to generate highly detailed SI depth maps and achieve generalization in diverse scenarios.
format Preprint
id arxiv_https___arxiv_org_abs_2406_09374
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Scale-Invariant Monocular Depth Estimation via SSI Depth
Miangoleh, S. Mahdi H.
Reddy, Mahesh
Aksoy, Yağız
Computer Vision and Pattern Recognition
I.4.5
Existing methods for scale-invariant monocular depth estimation (SI MDE) often struggle due to the complexity of the task, and limited and non-diverse datasets, hindering generalizability in real-world scenarios. This is while shift-and-scale-invariant (SSI) depth estimation, simplifying the task and enabling training with abundant stereo datasets achieves high performance. We present a novel approach that leverages SSI inputs to enhance SI depth estimation, streamlining the network's role and facilitating in-the-wild generalization for SI depth estimation while only using a synthetic dataset for training. Emphasizing the generation of high-resolution details, we introduce a novel sparse ordinal loss that substantially improves detail generation in SSI MDE, addressing critical limitations in existing approaches. Through in-the-wild qualitative examples and zero-shot evaluation we substantiate the practical utility of our approach in computational photography applications, showcasing its ability to generate highly detailed SI depth maps and achieve generalization in diverse scenarios.
title Scale-Invariant Monocular Depth Estimation via SSI Depth
topic Computer Vision and Pattern Recognition
I.4.5
url https://arxiv.org/abs/2406.09374