{"ID":2842445,"CreatedAt":"2026-06-01T04:54:23.091178241Z","UpdatedAt":"2026-06-01T04:54:23.091178241Z","DeletedAt":null,"paper_url":"https://arxiv.org/abs/2511.10647","arxiv_id":"2511.10647","title":"Depth Anything 3: Recovering the Visual Space from Any Views","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.","short_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 a...","url_abs":"https://arxiv.org/abs/2511.10647","url_pdf":"https://arxiv.org/pdf/2511.10647v1","authors":"[\"Haotong Lin\",\"Sili Chen\",\"Junhao Liew\",\"Donny Y. Chen\",\"Zhenyu Li\",\"Guang Shi\",\"Jiashi Feng\",\"Bingyi Kang\"]","published":"2025-11-13T18:59:53Z","proceeding":"cs.CV","tasks":"[\"cs.CV\"]","methods":"[\"Transformer\"]","has_code":false}