The Year in AI Papers
The year/Independent research

Paper 2511.10647

Depth Anything 3: Recovering the Visual Space from Any Views

Haotong Lin, Sili Chen, Junhao Liew, Donny Y. Chen, Zhenyu Li, Guang Shi, Jiashi Feng, Bingyi Kang

Published
Nov 2025
Research lab
Independent
Citations
499
GitHub
6.1K stars

01 In brief

Summary

Depth Anything 3 (DA3) is a model that predicts spatially consistent geometry from any number of images, with or without known camera poses.

It uses a single plain transformer (e.g., vanilla DINOv2) as backbone, with an input-adaptive cross-view self-attention mechanism and a dual-DPT head that jointly outputs depth and ray maps.

A depth-ray representation serves as the minimal prediction target, and camera poses can be optionally injected via a lightweight encoder.

Training follows a teacher-student paradigm: a teacher model trained on synthetic data generates pseudo-depth labels for real-world data, which are aligned to noisy ground truth.

DA3 sets a new state-of-the-art on a new visual geometry benchmark (HiRoom, ETH3D, DTU, 7Scenes, ScanNet++), surpassing VGGT by 35.7% in pose accuracy and 23.6% in geometric accuracy, and outperforms Depth Anything 2 in monocular depth estimation.

It also powers a feed-forward 3D Gaussian splatting model for novel view synthesis, outperforming specialized models.

All models are trained exclusively on public academic datasets.

02 From the paper

Abstract

We present Depth Anything 3 (DA3), a model that predicts spatially consistent geometry from an arbitrary number of visual inputs, with or without known camera poses. In pursuit of minimal modeling, DA3 yields two key insights: a single plain transformer (e.g., vanilla DINO encoder) is sufficient as a backbone without architectural specialization, and a singular depth-ray prediction target obviates the need for complex multi-task learning. Through our teacher-student training paradigm, the model achieves a level of detail and generalization on par with Depth Anything 2 (DA2). We establish a new visual geometry benchmark covering camera pose estimation, any-view geometry and visual rendering. On this benchmark, DA3 sets a new state-of-the-art across all tasks, surpassing prior SOTA VGGT by an average of 44.3% in camera pose accuracy and 25.1% in geometric accuracy. Moreover, it outperforms DA2 in monocular depth estimation. All models are trained exclusively on public academic datasets.