Single-Image Depth Perception in the Wild

By Weifeng Chen, Zhao Fu, Dawei Yang, Jia Deng. 2016

575 citations recorded by the source corpus.

This paper studies single-image depth perception in the wild, i.e., recovering depth from a single image taken in unconstrained settings. We introduce a new dataset "Depth in the Wild" consisting of images in the wild annotated with relative depth between pairs of random points. We also propose a new algorithm that learns to estimate metric depth using annotations of relative depth. Compared to the state of the art, our algorithm is simpler and performs better. Experiments show that our algorithm, combined with existing RGB-D data and our new relative depth annotations, significantly improves single-image depth perception in the wild.

Subfields: Computer Vision and Pattern Recognition, Media Technology

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Computer Vision and Pattern Recognition

In Media Technology

Citations in the corpus

Learning to Reconstruct Shapes from Unseen Classes (2018, background): “…RGB-D datasets become available [McCormac et al., 2017, Silberman et al., 2012, Song et al., 2017], many papers start to estimate depth [Chen et al., 2016, Eigen and Fergus, 2015], surface normals [Bansal and Russell, 2016, Wang et al., 2015], and other intrinsic images [Janner et al.,…”
Learning Shape Priors for Single-View 3D Completion and Reconstruction (2018, background): “With the development of depth sensors [23] and larger-scale RGB-D datasets [33,42,43], there have also been papers on estimating depth [6,10], surface normals [1,54], and other intrinsic images [25,41] with deep networks.”
MarrNet: 3D Shape Reconstruction via 2.5D Sketches (2017, background): “…and larger scale RGB-D datasets [Silberman et al., 2012, Song et al., 2017, McCormac et al., 2017], there have also been papers on estimating depth [Chen et al., 2016, Eigen and Fergus, 2015], surface normals [Bansal and Russell, 2016, Wang et al., 2015], and other intrinsic images [Shi et al.,…”
Towards Robust Monocular Depth Estimation: Mixing Datasets for Zero-Shot Cross-Dataset Transfer (2019, methodology): “B-D 2.5M Make3D [37] 3 3 3 Low Low Laser 534 KITTI LiDAR [14, 32] 3 3 3 3 Medium Low Laser 93K KITTI Stereo [14, 32] 3 3 3 3 3 Medium Low Stereo 93K Cityscapes [6] 3 3 3 3 3 Medium Low Stereo 25K DIW [3] 3 3 3 Low High User clicks 496K ETH3D [38] 3 3 3 3 High Low Laser 454 Tanks and Temples [24] 3 3”
Visual SLAM for Automated Driving: Exploring the Applications of Deep Learning (2018, background): “Interestingly, [11] uses a ranking loss that penalizes the non relative correspondence between predictions and ground truth while [4] defines depth estimation as a classification problem.”
Repurposing Diffusion-Based Image Generators for Monocular Depth Estimation (2023, unknown): “DIW [8] was perhaps the earliest work to introduce an uncontrolled dataset and to predict relative (ordinal) depth for it.”
Clear Grasp: 3D Shape Estimation of Transparent Objects for Manipulation (2019, background): “More recent works study how to obtain 3D data from color images by directly inferring depth images [14, 29, 10, 40, 22, 53, 16, 45], or filling in missing depth values in RGB-D images captured by commodity 3D cameras [21, 17, 63].”
Consistent video depth estimation (2020, methodology): “2016a] followed by domain adaptation [Atapour-Abarghouei and Breckon 2018], collecting relative depth annotations [Chen et al. 2016], using conventional structure-from-motion and multi-view stereo algorithms to obtain pseudo ground truth depth maps from Internet images [Chen et al.”
DF-Net: Unsupervised Joint Learning of Depth and Flow using Cross-Task Consistency (2018, background): “Some efforts [8,74] have been made to relax the difficulty of collecting absolute depth by exploring learning from relative/ordinal depth annotations.”
Depth Anything V2 (2024, unknown): “Although DIW [11] and our DA-2K use the same annotation format (both sparse pixel pairs, we are inspired by DIW), our proposed DA-2K dataset is better in four aspects: • (more precise) DIW is very noisy.”
Deep Ordinal Regression Network for Monocular Depth Estimation (2018, methodology): “Also, some weakly-supervised methods considering pair-wise ranking information were proposed to roughly estimate and compare depth [66, 7].”
BLINK: Multimodal Large Language Models Can See but Not Perceive (2024, methodology): “Part of our samples are curated from [10,19,29,31,35,42,88].”
Semi-Supervised Deep Learning for Monocular Depth Map Prediction (2017, background): “[1] incorporates the side-task of depth ranking of pairs of pixels for training a CNN on single image depth prediction.”
Neural Style Transfer: A Review (2017, unknown): “’s work in [64]).”
SparseNeRF: Distilling Depth Ranking for Few-shot Novel View Synthesis (2023, background): “, user clicks [6], RGB-D [34, 37], and laser/stereo [25]), dataset bias, and imperfect depth estimation models, it is challenging to obtain accurate 3D depth estimation given 2D singleview images.”
Defocus Blur Detection via Depth Distillation (2020, methodology): “[3] propose an end-to-end network based on point relationships.”
Metric3D: Towards Zero-shot Metric 3D Prediction from A Single Image (2023, methodology): “According to the predicted depth, existing methods are categorized into learning metric depth [71, 64, 4, 63], learning relative depth [57, 58, 8, 7], and learning affine-invariant depth [69, 68, 40, 39, 73].”
SynSin: End-to-End View Synthesis From a Single Image (2019, background): “Many works explore using a DNN to predict 3D object shapes [18, 20, 24, 26, 64, 69] or the depth of a scene given an image [6, 13, 33, 73].”
3D Photography Using Context-Aware Layered Depth Inpainting (2020, background): “While the accuracy of these approaches is not yet competitive with multi-view stereo algorithms, this line of research is particularly promising due to the availability of larger and more diverse training datasets from relative depth annotations [3], multi-view stereo [30], 3D movies [28] and synthe”
Depth-Aware Video Frame Interpolation (2019, methodology): “We use the hourglass architecture [6] as our depth estimation network.”