# How do neural networks see depth in single images?

arXiv: https://arxiv.org/abs/1905.07005 · HTML: https://mlsanity.com/paper/1905.07005/

## Key facts

- Cited by 82 papers in the map (data through 14 Oct 2026).
- First posted on arXiv 16 May 2019.

## Built on by (most cited first, 30 of 82)

- [Geometry Uncertainty Projection Network for Monocular 3D Object Detection](https://mlsanity.com/paper/2107.13774/) (2021)
- [Multi-Frame Self-Supervised Depth with Transformers](https://mlsanity.com/paper/2204.07616/) (2022)
- [Global-Local Path Networks for Monocular Depth Estimation with Vertical CutDepth](https://mlsanity.com/paper/2201.07436/) (2022)
- [STS: Surround-view Temporal Stereo for Multi-view 3D Detection](https://mlsanity.com/paper/2208.10145/) (2022)
- [Virtual Normal: Enforcing Geometric Constraints for Accurate and Robust Depth Prediction](https://mlsanity.com/paper/2103.04216/) (2021)
- [Excavating the Potential Capacity of Self-Supervised Monocular Depth Estimation](https://mlsanity.com/paper/2109.12484/) (2021)
- [Self-supervised Monocular Depth Estimation for All Day Images using Domain Separation](https://mlsanity.com/paper/2108.07628/) (2021)
- [Toward Practical Monocular Indoor Depth Estimation](https://mlsanity.com/paper/2112.02306/) (2021)
- [Depth Any Camera: Zero-Shot Metric Depth Estimation from Any Camera](https://mlsanity.com/paper/2501.02464/) (2025)
- [DID-M3D: Decoupling Instance Depth for Monocular 3D Object Detection](https://mlsanity.com/paper/2207.08531/) (2022)
- [MonoCD: Monocular 3D Object Detection with Complementary Depths](https://mlsanity.com/paper/2404.03181/) (2024)
- [Object as Query: Lifting any 2D Object Detector to 3D Detection](https://mlsanity.com/paper/2301.02364/) (2023)
- [Single Image Depth Estimation: An Overview](https://mlsanity.com/paper/2104.06456/) (2021)
- [Graph-DETR3D: Rethinking Overlapping Regions for Multi-View 3D Object Detection](https://mlsanity.com/paper/2204.11582/) (2022)
- [YOLOStereo3D: A Step Back to 2D for Efficient Stereo 3D Detection](https://mlsanity.com/paper/2103.09422/) (2021)
- [Self-supervised Monocular Depth Estimation: Let's Talk About The Weather](https://mlsanity.com/paper/2307.08357/) (2023)
- [Deep Depth Completion from Extremely Sparse Data: A Survey](https://mlsanity.com/paper/2205.05335/) (2022)
- [PlaneDepth: Self-supervised Depth Estimation via Orthogonal Planes](https://mlsanity.com/paper/2210.01612/) (2022)
- [GEDepth: Ground Embedding for Monocular Depth Estimation](https://mlsanity.com/paper/2309.09975/) (2023)
- [Depth Prompting for Sensor-Agnostic Depth Estimation](https://mlsanity.com/paper/2405.11867/) (2024)
- [Near-field Perception for Low-Speed Vehicle Automation using Surround-view Fisheye Cameras](https://mlsanity.com/paper/2103.17001/) (2021)
- [MUAD: Multiple Uncertainties for Autonomous Driving, a benchmark for multiple uncertainty types and tasks](https://mlsanity.com/paper/2203.01437/) (2022)
- [MonoDGP: Monocular 3D Object Detection with Decoupled-Query and Geometry-Error Priors](https://mlsanity.com/paper/2410.19590/) (2024)
- [Learning Depth via Leveraging Semantics: Self-supervised Monocular Depth Estimation with Both Implicit and Explicit Semantic Guidance](https://mlsanity.com/paper/2102.06685/) (2021)
- [Exploring Geometric Consistency for Monocular 3D Object Detection](https://mlsanity.com/paper/2104.05858/) (2021)
- [Stereoscopic Universal Perturbations across Different Architectures and Datasets](https://mlsanity.com/paper/2112.06116/) (2021)
- [Latent Discriminant deterministic Uncertainty](https://mlsanity.com/paper/2207.10130/) (2022)
- [LightDepth: Single-View Depth Self-Supervision from Illumination Decline](https://mlsanity.com/paper/2308.10525/) (2023)
- [URCDC-Depth: Uncertainty Rectified Cross-Distillation with CutFlip for Monocular Depth Estimation](https://mlsanity.com/paper/2302.08149/) (2023)
- [Lift3D: Synthesize 3D Training Data by Lifting 2D GAN to 3D Generative Radiance Field](https://mlsanity.com/paper/2304.03526/) (2023)

## Abstract (arXiv)

Deep neural networks have lead to a breakthrough in depth estimation from single images. Recent work often focuses on the accuracy of the depth map, where an evaluation on a publicly available test set such as the KITTI vision benchmark is often the main result of the article. While such an evaluation shows how well neural networks can estimate depth, it does not show how they do this. To the best of our knowledge, no work currently exists that analyzes what these networks have learned. In this work we take the MonoDepth network by Godard et al. and investigate what visual cues it exploits for depth estimation. We find that the network ignores the apparent size of known obstacles in favor of their vertical position in the image. Using the vertical position requires the camera pose to be known; however we find that MonoDepth only partially corrects for changes in camera pitch and roll and that these influence the estimated depth towards obstacles. We further show that MonoDepth's use of the vertical image position allows it to estimate the distance towards arbitrary obstacles, even those not appearing in the training set, but that it requires a strong edge at the ground contact point of the object to do so. In future work we will investigate whether these observations also apply to other neural networks for monocular depth estimation.