Roboflow

Depth Estimation Models

The one depth estimation model in our catalog.

1 model · 1 open-weight · prices synced Oct 27, 2026

Open-Source Depth Estimation Models

1 model with downloadable weights you can self-host under their licenses (Apache 2.0).

ByteDance
Depth Anything V2
Depth Anything V2 is a monocular depth estimation model released in June 2024 by researchers at the University of Hong Kong and TikTok. It predicts a dense depth map from a single RGB image, enabling 3D-aware applications without the need for stereo cameras, LiDAR, or multi-view inputs. The model improves on the original Depth Anything through three modifications: replacing real labeled images with 595K high-quality synthetic images during teacher training, scaling up teacher model capacity, and using the stronger teacher to generate pseudo-labels on 62 million unlabeled real images used to train the student models. This pipeline reduces the depth prediction artifacts that can occur in reflective, transparent, and texture-poor regions. Compared to diffusion-based depth models such as Marigold, Depth Anything V2 runs more than 10× faster while producing more accurate predictions.Depth Anything V2 is released in four sizes: Small (25M), Base (97M), Large (335M), and Giant (1.3B), and in two output modes: relative depth (normalized scene-level estimates) and metric depth (absolute distance in meters, produced by fine-tuning the relative-depth backbone on depth-annotated datasets). The Small, Base, and Large model weights are released under Apache 2.0, and the Giant variant under CC-BY-NC-4.0 for non-commercial use. A successor model, Depth Anything 3, was released in November 2025 by the ByteDance Seed team, extending the framework to multi-view depth estimation and camera pose recovery.

How to Choose a Depth Estimation Model

Depth models predict distance per pixel, and the deciding question is whether you need relative depth (what is nearer) or metric depth (how far, in units), because those are different problems.

Monocular models in the MiDaS and Depth Anything lineage generalize across scene types and produce excellent relative depth maps: correct ordering and clean boundaries. That is enough for background blur, AR occlusion, image editing, and scene understanding, and it works with any ordinary camera.

Real distances require metric depth models (trained for absolute scale, often per-domain), stereo cameras, or depth sensors. Monocular metric estimates carry real error that varies by scene, so navigation, measurement, and safety systems should verify against ground truth in their own environment before trusting any model output.

The bottom line: For ordering and visual effects, a monocular relative-depth model is enough. For real distances, use a metric model validated in your environment, or add stereo or a sensor.

Explore other vision task pages, compare models head to head, or browse the full model catalog.