← Back to Model Beat
Research·4d ago·all news from November 18, 2026

Depth Pro 2: Sharp and Temporally Consistent Metric Monocular Depth

Apple has introduced Depth Pro 2, a method designed to estimate dense metric depth from images and video at full resolution. The model addresses a common technical trade-off by improving both boundary sharpness and temporal consistency across frames. This development enhances the reliability of metric-scale geometry recovery, which is critical for applications like high-quality novel view synthesis. By refining how monocular systems interpret spatial depth, the technology enables more stable and accurate 3D reconstructions from standard visual input.

Covered by 2 sources · 5 articles

  • AApple Machine Learning Blog↗2d ago
  • AarXiv CS.AI↗Peiyun Hu, Marko Mihajlovic, Lyndon Duong, Lahav Lipson, Wei Dong, Ama\"el Delaunoy, Stephan R. Richter, Vladlen Koltun2d ago
  • AarXiv CS.AI↗Shiori Ueda, Rinto Yagawa, Hideo Saito, Shohei Mori2d ago
  • AarXiv CS.AI↗Rizhao Fan, Zhigen Li, Heping Li, Ning An4d ago
  • AarXiv CS.AI↗Luosong Guo, Xiang Yang, Chaoyang Li, Ao Gao, Jiangming Shi, Yanyun Qu, Zhizhong Zhang, Jingyu Gong, Yuan Xie3d ago

Related stories

ResearchResearch acceleration: The jagged frontierNov 17 · 4 sourcesResearchOur first phase of embedded assessmentsNov 16 · 4 sourcesResearchTop AI Models Regularly Identify Civilians as Military Targets in TestsNov 17 · 4 sourcesResearchGoogle, Salesforce, Sarvam AI partner with Karnataka to train 63,200 people in AINov 17 · 12 sources