Apple releases Depth Pro 2 models for sharp, steady depth from photos and video
Apple has released Depth Pro 2, a family of models that estimate metric depth for every pixel from an ordinary camera, whether a single photo, a whole video or a live feed, at the input's full resolution, according to its Hugging Face model card.
The models come in four sizes, from 24 million to 872 million parameters, each in its own repository, and share a 23.5 million parameter estimator that predicts focal length when it is unknown. Larger sizes give sharper and steadier depth, while smaller ones run faster. A multi-view transformer processes one or more frames jointly, alternating attention within and across frames, and a decoder predicts a depth map for each frame.
Apple says Depth Pro 2 outperforms prior single-frame and multi-frame methods on edge sharpness and frame-to-frame consistency, with results averaged over the Sintel and Spring datasets. The weights are released under the Apple Machine Learning Research Model licence for non-commercial research use only. The code is on GitHub, and an arXiv paper by Peiyun Hu, Vladlen Koltun and colleagues describes the method.
We hope you enjoyed this article
Consider subscribing to one of our newsletters like Daily AI Brief.
Also, consider following us on social media:
Daily AI Brief
Daily report covering major AI developments and industry news, with both top stories and complete market updates
Whitepaper
Tensordyne Napier: What If One Rack Could Do the Work of Nine?
Tensordyne
This Tensordyne whitepaper presents Napier, an inference-focused AI processor and rack-scale system based on the company’s TDN Math logarithmic number system. It examines infrastructure requirements for large mixture-of-experts and agentic models, compares major inference architecture approaches, and details the TDN AIP processor, TDN72 pod, TDN Link fabric, and Napier Ultra configuration. The paper reports simulation-based performance, cost, and accuracy-validation results, including Tensordyne’s projected comparison of one Napier rack with a nine-rack Nvidia Rubin plus Groq deployment; the chip is reported as taped out and in fabrication.
Read moreYou may also like
Liquid AI Releases LFM3, Six Open-Weight Models for On-Device AI
Adobe previews precise AI editing projects at MAX
MiniMax Releases Open-Weight H3-Image Model for 4K Images in Under 10 Seconds
DeepX tapes out 2nm DX-M2 edge AI chip built for 100B-parameter models at 5W
Deep Principle releases Hydrogen-1.0 scientific model weights
Daily AI Brief: the AI news that matters, in your inbox.