Single Image Depth Estimation: An Overview

Mertan, Alican; Duff, Damien Jade; Unal, Gozde

doi:10.48550/arxiv.2104.06456

Cited by 4 publications

(3 citation statements)

References 57 publications

(146 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In a survey regarding single image depth estimation, ref. [42] present a broad range of state-of-the-art methods which also cover the category of self-supervised joint learning of depth end ego-motion.…”

Section: Learning Depth and Ego-motionmentioning

confidence: 99%

Self-Supervised Steering and Path Labeling for Autonomous Driving

Mihalea,

Samoilescu,

Florea

2023

Sensors

View full text Add to dashboard Cite

Autonomous driving is a complex task that requires high-level hierarchical reasoning. Various solutions based on hand-crafted rules, multi-modal systems, or end-to-end learning have been proposed over time but are not quite ready to deliver the accuracy and safety necessary for real-world urban autonomous driving. Those methods require expensive hardware for data collection or environmental perception and are sensitive to distribution shifts, making large-scale adoption impractical. We present an approach that solely uses monocular camera inputs to generate valuable data without any supervision. Our main contributions involve a mechanism that can provide steering data annotations starting from unlabeled data alongside a different pipeline that generates path labels in a completely self-supervised manner. Thus, our method represents a natural step towards leveraging the large amounts of available online data ensuring the complexity and the diversity required to learn a robust autonomous driving policy.

show abstract

Section: Learning Depth and Ego-motionmentioning

confidence: 99%

Self-Supervised Steering and Path Labeling for Autonomous Driving

Mihalea,

Samoilescu,

Florea

2023

Sensors

View full text Add to dashboard Cite

show abstract

“…Single image depth estimation [12] aims at estimating a dense depth map from a single RGB image. Occlusion between objects and perspective, including size cue and texture gradient, are keys for monocular depth estimation [27], and several learning models [11,14,4] based on these cues are proposed. Apart from learning the depth estimation individually, some works jointly solve single image depth estimation task with other similar tasks such as semantic segmentation [20], surface normal estimation [31] and contour estimation [43].…”

Section: Single Image Depth Estimationmentioning

confidence: 99%

Spatial and Visual Perspective-Taking via View Rotation and Relation Reasoning for Embodied Reference Understanding

Shi

Yang

2022

Lecture Notes in Computer Science

View full text Add to dashboard Cite

Embodied Reference Understanding studies the reference understanding in an embodied fashion, where a receiver requires to locate a target object referred to by both language and gesture of the sender in a shared physical environment. Its main challenge lies in how to make the receiver with the egocentric view access spatial and visual information relative to the sender to judge how objects are oriented around and seen from the sender, i.e., spatial and visual perspective-taking. In this paper, we propose a REasoning from your Perspective (REP) method to tackle the challenge by modeling relations between the receiver and the sender as well as the sender and the objects via the proposed novel view rotation and relation reasoning. Specifically, view rotation first rotates the receiver to the position of the sender by constructing an embodied 3D coordinate system with the position of the sender as the origin. Then, it changes the orientation of the receiver to the orientation of the sender by encoding the body orientation and gesture of the sender. Relation reasoning models both the nonverbal and verbal relations between the sender and the objects by multi-modal cooperative reasoning in gesture, language, visual content, and spatial position. Experiment results demonstrate the effectiveness of REP, which consistently surpasses all existing state-of-the-art algorithms by a large margin, i.e., +5.

show abstract

“…Since this usually involves the derivation of distances between the navigation subject and the environment (i.e., spatial depth), this research field is referred to in the literature as single image depth estimation (SIDE). For a summary of the developments in the general SIDE field, which mainly refers to the analysis of classical optical images with horizontal viewing direction, interested readers are referred to the overview article by (Mertan et al, 2021).…”

Section: Introductionmentioning

confidence: 99%

Comparison of Single-Image Urban Height Reconstruction From Optical and Sar Data

Schmitt

Recla

2022

Int. Arch. Photogramm. Remote Sens. Spatial Inf. Sci.

View full text Add to dashboard Cite

Abstract. Deep learning-based depth estimation has become an important topic in recent years, not only in the field of computer vision. Also in the context of remote sensing, scientists started a few years ago to adapt or develop suitable approaches to realize a reconstruction of the Earth’s surface without requiring several images. There are many reasons for this: First, of course, the aspect of general economization, since especially high-resolution satellite images are often accompanied by high data acquisition costs. In addition, there is also the desire to be able to acquire high-quality geoinformation as quickly as possible in time-critical cases – for example, the provision of up-to-date maps for emergency forces in disaster scenarios. Finally, a reconstruction of topography based only on single images can also provide important approximate values for the classic multi-image methods. For example, various processing steps in a classical InSAR process chain require a rough knowledge of the Earth’s surface in order to achieve the most accurate and reliable results. In this paper, we review the developments documented in the remote sensing literature so far. Using an established neural network architecture, we produce example results for both very-high-resolution SAR and optical imagery. The comparison shows that SAR-based single-image-height reconstruction seems to bear an even greater potential than single-image height reconstruction from optical data.

show abstract

Single Image Depth Estimation: An Overview

Cited by 4 publications

References 57 publications

Self-Supervised Steering and Path Labeling for Autonomous Driving

Self-Supervised Steering and Path Labeling for Autonomous Driving

Spatial and Visual Perspective-Taking via View Rotation and Relation Reasoning for Embodied Reference Understanding

Comparison of Single-Image Urban Height Reconstruction From Optical and Sar Data

Contact Info

Product

Resources

About