Monocular depth
Estimating object distances in a scene from a single camera image.
Monocular depth is a computer vision technique that allows an AI to guess how far away objects are in a picture, using only a single camera image. It's like how humans can often tell if something is close or far just by looking at a photo.
For example, if a photo shows a person standing in front of a building, monocular depth estimation would try to figure out the distance to the person and the distance to the building. This helps AI understand the 3D layout of a scene from a 2D image.
Monocular depth estimation infers scene geometry, specifically the distance of points from the camera, using a single 2D image. It typically relies on deep learning models, often convolutional neural networks (CNNs) or transformers, trained on large datasets of images paired with ground truth depth information (e.g., from LiDAR or stereo cameras). The models learn to associate visual cues like object size, occlusion, texture gradients, and perspective with depth. Tradeoffs include potential inaccuracies in complex scenes, reliance on training data distribution, and difficulty with reflective or transparent surfaces.
This capability is becoming more prominent as AI models improve their understanding of visual scenes. Google's use of it in photo editing shows how monocular depth estimation can enable new creative tools by reconstructing scene geometry from static images.
- It's all about the angle: Your photos, re-composedGoogle · Google Research Blog ·
- Google Gemini
- Google uses monocular depth estimation to power features like re-framing photos from new angles, demonstrating its application in creative visual editing tools.
- Stereo depth
- This method uses two cameras to calculate depth, similar to human
- Structure from motion
- This technique estimates both camera motion and 3D struct
The concept of inferring depth from monocular cues has a long history in computer vision and psychology. Deep learning approaches gained significant traction in the mid-2010s with the advent of large-scale datasets and powerful CNN architectures. Source ↗
Take a photo with your smartphone. Then, try using a photo editing app that offers features like object removal or background blur. The app likely uses monocular depth estimation to understand which parts of the image are foreground and background, allowing it to manipulate them separately.
The Playbook: reusable how-tos, prompts, configs and tips pulled out of the news.
- PromptRole-play a student struggling with math for AI tutor simulationSimulate a student's math struggles and uncertainty for AI tutor evaluation. · 5 min
- TemplateReusable workflow templates for healthcare teamsTeams can create and share templates for common workflows like documentation or prior authorizations. · 15 min
- TemplateTemplate for building an AI tool tailored to your workA structured approach to defining and building a custom AI tool for specific business needs. · 15 min
- TemplateEvals CSV prompt list for agent skillsA CSV file template for listing prompts used in evaluating agent skills. · 20 min
- HeuristicDouble coding agent task completion with docs linksProviding coding agents with a documentation link significantly increases their task completion rate. · 2 min
Questions people ask
- How accurate is monocular depth estimation?
- Accuracy varies greatly depending on the model, training data, and scene complexity. It is generally less accurate than methods using multiple cameras or sensors like LiDAR, but is rapidly improving.
- What visual cues do AI models use for depth?
- Models learn from cues like object size (closer objects appear larger), occlusion (objects blocking others are closer), texture gradients (textures appear finer further away), and perspective.
First seen on Subvolts Apr 22, 2026. Updated Sep 15, 2026. Written by an AI model from the items above and reviewed against them weekly; report an error.