微软开源的统一 3D 生成模型，用结构化 3D 潜变量把文本或图像转成高质量 3D 资产，支持大规模、多用途的 3D 内容创作。
开源多模态大模型家族，视觉语言能力接近 GPT-4o，支持图像分类、图文检索、语义分割和视频理解等任务。
汇总 CVPR 2026 录用论文及其配套开源代码，方便计算机视觉研究者快速找到带实现的新工作并跟进复现。
MoGe-2 ViT-B + normal (CVPR 2025 Oral) — monocular 3D geometry estimation. Single CoreML mlpackage (~200 MB FP16, 504×504 fixed input). Predicts metric depth, surface nor
…distance metric ( 1246) - VisRes Bench benchmark (CVPR 2026) ( 1245) Fixes - Cross-class filelock deadlock in datasets loading ( 1253) - CharXiv lazy-init OpenAI client a
…Roth, 2024](https://openaccess.thecvf.com/content/CVPR2024/papers/Steitz Adapters Strike Back CVPR 2024 paper.pdf)) is a new bottleneck adapter variant optimized for visi
…ng data, and evaluatiton. - Model weights for our CVPR models. - Model weights after upgrading software and cleaning code for the repository. Have fun folks!
…r](https://arxiv.org/abs/2012.09988) to appear in CVPR-2021. - Python and Web API for Objectron models (available via [Mediapipe](https://google.github.io/mediapipe/solut
… a RegionMatcherFactory 1056 Multiview: - Fix the CVPR 2017 resection solver 1253 - Add an orthographic essential matrix solver "Two-View Orthographic Epipolar Geometry: