README.md

---
title: Depth Anything V2
emoji: 🌖
colorFrom: indigo
colorTo: purple
sdk: static
app_file: index.html
short_description: In-browser monocular depth estimation with Transformers.js
license: apache-2.0
---

# Depth Anything V2 — In-Browser Demo

Runs **Depth Anything V2** (Yang et al., 2024 — [arXiv 2406.09463](https://arxiv.org/abs/2406.09463))
entirely in your browser via [Transformers.js](https://huggingface.co/docs/transformers.js) and ONNX Runtime.
No server, no GPU — all inference happens client-side on WebAssembly / WebGPU.

## Models

Uses the ONNX ports from `onnx-community`:

| Option | Model repo | Params |
|--------|-----------|--------|
| Small  | `onnx-community/depth-anything-v2-small-ONNX`  | 24.8M |
| Base   | `onnx-community/depth-anything-v2-base-ONNX`   | 97.5M |
| Large  | `onnx-community/depth-anything-v2-large-ONNX`  | 335.3M |

## How it works

1. Upload an image (or pick an example).
2. Select a model size.
3. Click **Compute Depth** — the model downloads (cached afterwards) and runs locally.
4. Compare the original and depth map with the slider.