Depth Anything V2 Small

Depth-Anything-V2-Metric-Indoor-Small-hf is a monocular depth estimation model fine-tuned for indoor metric depth prediction. It is built on the Depth Anything V2 architecture, which uses a DPT architecture with a DINOv2 backbone and 24.8M parameters.

The model is fine-tuned on the synthetic Hypersim dataset to produce metric depth values for indoor scenes. Depth Anything V2 was trained on approximately 600K synthetic labeled images and 62 million real unlabeled images, achieving state-of-the-art results for both relative and absolute depth estimation. Compared to the original Depth Anything V1, this version produces finer and more robust depth predictions by replacing all labeled real images with synthetic images, scaling up the teacher model capacity, and teaching student models via large-scale pseudo-labeled real images.

Key strengths include significantly finer and more robust depth predictions than V1, and substantially faster inference (more than 10x) compared to models built on Stable Diffusion while maintaining higher accuracy. The model is part of a family of six metric depth models across three scales (Small, Base, Large) for indoor and outdoor scenes.

The model is hosted on gigarouter as a managed, OpenAI-compatible API — no local installation or dependency management is required.