Depth Anything V2 Small
depth-anything/Depth-Anything-V2-Small
published Jun 2024 · updated Jul 2024
Depth Anything V2 Small is a monocular depth estimation model that produces fine-grained, robust depth maps from a single image.
specs
| Task | Monocular Depth Estimation |
| Architecture | Vision Transformer Small (ViT-S) |
| Parameters | 25 million |
| Training Data | 595K synthetic labeled images + 62M+ real unlabeled images |
about this model
Depth-Anything-V2-Small is a monocular depth estimation model that produces dense depth maps from a single RGB image. It is the small variant (25M parameters, ViT-S encoder) of the Depth Anything V2 family, trained on 595K synthetic labeled images and over 62 million real unlabeled images.
Compared with its predecessor Depth Anything V1, this version delivers finer-grained details and more robust depth predictions. Against Stable Diffusion–based models such as Marigold and GeoWizard, it offers superior robustness while being more than 10x faster and significantly more lightweight. The model’s strong generalization enables effective fine-tuning for metric depth estimation tasks.
Key attributes:
- Fine-grained and robust monocular depth estimation
- 10x faster inference than SD-based alternatives
- Pretrained on synthetic data and scaled via pseudo-labeled real images
On gigarouter, Depth-Anything-V2-Small is available as a managed, OpenAI-compatible API—no installation or model loading required.
best for
- ·Real-time depth estimation on edge devices
- ·Robust depth prediction for robotics and autonomous navigation
- ·High-quality depth maps for 3D reconstruction
FAQ
The model takes a single RGB image as input and outputs a raw depth map of the same spatial size (HxW).
V2 produces finer details, is more robust, and is more efficient than V1. It is also 10x faster than Stable Diffusion-based depth models.
Yes, the pre-trained model can be fine-tuned with metric depth labels to obtain metric depth predictions.
Use the OpenAI-compatible endpoint with your API key, sending an image URL or base64 in the request.
The license is not explicitly stated in the model card; please check the official repository for details.
We're benchmarking and onboarding Depth Anything V2 Small as a hosted, OpenAI-compatible API. Sign in for free credit and be ready when it lands, or tell us you want it and we'll prioritize it.