YOLO-StereoDepth is an announced stereo depth estimation model in the YOLO family, computing metric depth from two cameras using binocular disparity, positioned as a camera-native alternative to lidar for robotics, expected September 2026.
YOLO models are a family of real-time computer vision models designed to handle a wide range of tasks, including object detection, segmentation, pose estimation, and classification.
YOLO-StereoDepth has been announced as part of YOLO27, described as binocular disparity depth for robotics, a camera-native alternative to lidar, with a release planned for September 2026.
In this blog, we'll cover what YOLO-StereoDepth is, what it could be used for, how stereo compares to monocular depth and when to use which, what remains unknown ahead of release, and how to read metric depth per detection on edge hardware today.
What Is YOLO-StereoDepth?
YOLO-StereoDepth is an announced stereo depth estimation model in the YOLO family, unveiled as part of the YOLO27 generation alongside its monocular sibling, YOLO-Depth.
Stereo depth uses binocular disparity, the same principle as human vision. Two cameras a known distance apart (the baseline) capture the same scene, and the model computes depth from the difference between the two views. Because the baseline is known, stereo depth produces absolute, metric distances: 1.42 meters, not "closer than the shelf behind it."
That is the key difference from monocular depth, and it is why the announcement aims this model at robotics. A robot acting in the physical world needs real units. It is also why the announcement frames YOLO-StereoDepth as a camera-native alternative to lidar: a calibrated pair of commodity cameras costs a fraction of a lidar unit and captures color and texture lidar cannot.
Lidar keeps advantages in low light, low-texture scenes, and long range, so stereo will not replace it everywhere. For cost-sensitive robots working at room-to-warehouse distances, it does not need to.
What Could YOLO-StereoDepth Be Used For?
Stereo depth fits anywhere a machine needs metric distance to act:
- Robot navigation and obstacle stopping: an AMR needs to know it has 2.8 meters to stop, in meters, not in relative depth
- Grasping and bin picking: approach distance to the part, in real units, from the same sensor that detects the part
- Palletizing and depalletizing: layer heights and box positions for a robot arm working a pallet
- Dimensioning: measuring packages, parts, and loads in motion without a dimensioning tunnel
- Docking and alignment: vehicles and robots closing final distances against ramps, chargers, and conveyors
Monocular vs. Stereo Depth: When to Use Which
YOLO-Depth and YOLO-StereoDepth were announced together because they answer different deployment questions. The same is true of the options available today, monocular models like Depth Anything 3 versus stereo cameras. Here is how they compare:
| Monocular (one camera) | Stereo (two cameras) | |
|---|---|---|
| Depth type | Relative by default; metric requires calibration | Metric out of the box, from the known baseline |
| Hardware | Any existing RGB camera, no new capex | Stereo camera or calibrated pair, typically a few hundred dollars per unit |
| Accuracy | Consistent ordering of near and far; absolute error grows without calibration | Strong at short-to-mid range; error grows with distance squared as disparity shrinks |
| Weak spots | Unusual scenes that defeat learned priors; absolute scale | Textureless surfaces (blank walls, glass) where matching fails; needs good calibration |
| Compute | Neural depth model per frame on host or accelerator | Often computed on-camera, freeing host compute for detection |
| Best for | Retrofits on installed cameras, proximity ranking, monitoring and alerts | Robotics, grasping, dimensioning, anything that acts on real units |
The short version: if the cameras are already on the wall and you need to know what is closer to what, monocular depth gets you there without new hardware. If a machine has to move, grasp, or measure based on the number, stereo earns its hardware cost. Many operations end up with both, monocular on the installed fleet and stereo on the robots.
What We Don't Know Yet About YOLO-StereoDepth
As of this writing, we have not seen:
- Benchmarks: no accuracy or latency numbers, and no comparison against existing stereo matching methods or the onboard depth of current stereo cameras
- Camera support: whether YOLO-StereoDepth expects a specific calibrated rig, supports common stereo cameras out of the box, or takes any synchronized pair
- Baseline flexibility: how it handles different camera separations, which determine the usable depth range
- Model sizes and edge performance: stereo matching is compute-heavy, and whether this runs at frame rate on embedded robotics hardware is unstated
- Licensing: YOLO-StereoDepth licensing terms have not been announced. Previous similar releases shipped under AGPL-3.0, which requires open-sourcing derivative works unless you purchase a commercial license. Robotics deployments are almost always commercial, so this is worth confirming before you build on it.
- A paper: there are no indicated plans for a formal research paper
How to Read Metric Depth Per Detection on Edge Hardware Today
The combination YOLO-StereoDepth promises, real-time detection plus metric depth, is something you can deploy now using a stereo depth camera and Roboflow Inference on an edge device.
Stereo cameras like the Luxonis OAK-D, Stereolabs ZED, and Intel RealSense compute metric depth onboard and output an aligned depth frame alongside the RGB image. That onboard depth changes the engineering: no calibration workaround, no separate depth model competing for compute, just a distance value for any pixel you ask about.
The pattern:
- Run a detector on the RGB stream with Inference on your edge device. We recommend RF-DETR, trained on your own classes, and Roboflow supports a range of edge hardware for deployment.
- For each detection, sample the camera's aligned depth frame at the bounding box center (or the median over the box, which is more robust to edges) to get metric distance per object.
- Build the logic in Workflows: process the video stream, attach depth to every detection, and trigger actions, a stop signal when an obstacle is inside stopping distance, a grasp target for an arm, a dimension estimate for a passing package.
Because the camera supplies the depth and the detector supplies the what, you get YOLO-StereoDepth's headline capability, metric depth per detection from cameras, with hardware you can buy today.
RF-DETR is faster and more accurate than YOLO26 for object detection, and it ships with commercial-safe licensing. On a robot, the detector's speed budget matters twice: it shares the embedded compute with everything else, and its boxes decide which depth values the robot acts on.
YOLO-StereoDepth Alternatives
While YOLO-StereoDepth is not yet available, the detection-plus-metric-depth stack is deployable today.
Stereo Depth Cameras with RF-DETR
The pattern from the section above. This is the closest thing to YOLO-StereoDepth available now, and it will remain the baseline the new model has to beat.
Depth Anything 3
Depth Anything 3 is the current standard for monocular depth estimation and runs in the Roboflow Workflows Depth Estimation block. If your cameras are already installed and relative depth with a calibration step is enough, it covers the use cases that do not justify stereo hardware.
YOLO-Depth
Announced alongside YOLO-StereoDepth in the YOLO27 generation, YOLO-Depth is the monocular sibling: depth from a single camera, also expected September 2026. The comparison table above applies to the pair directly.
YOLO-StereoDepth Conclusion
YOLO-StereoDepth is the most robotics-specific model in the YOLO27 announcement: metric depth from commodity cameras, aimed at the gap between a $50 webcam and a lidar unit. For AMRs, arms, and automation at human scale, that gap is where most of the market lives.
Cite this Post
Use the following entry to cite this post in your research:
Contributing Writer. (May 28, 2026). What Is YOLO-StereoDepth?. Roboflow Blog: https://blog.roboflow.com/what-is-yolo-stereodepth/