magazine
2024.10.08

Apple's AI 'Depth Pro' Revolutionizes the World of New 3D Depth Recognition | Release #335

2024-10-apple-depth-pro-3d-cover-image

Cover photo by テル

Apple has announced its latest AI model, 'Depth Pro'. This is an astonishing technology that can generate a high-definition 3D depth map from a single 2D image in just 0.3 seconds.

While it is anticipated whether this technology will be installed in the latest iPhone, its applications are expected to extend to photo editing, AR, VR, and even autonomous driving. This time, we introduce the mechanism of 'Depth Pro' and its use cases.

Mechanism of Fast and Precise 3D Map Generation

'Depth Pro' adopts an efficient 'multi-scale vision transformer' design, capable of simultaneously grasping the overall structure and details of an image. Typically, depth estimation requires multiple images or camera setting information, but 'Depth Pro' eliminates this need, instantly calculating precise 3D data from a single 2D image.

2024-10-apple-depth-pro-3d-image-4
2024-10-apple-depth-pro-3d-image-5

Generated by the editorial team using the publicly available model

Additionally, by accurately tracing the contours of objects, it can faithfully reproduce fine structures such as hair and foliage. This enables realistic 3D experiences powered by AI, with new applications anticipated in AR and autonomous driving fields.

More Precise Bokeh and Subject Separation

By utilizing the high-definition depth maps generated by 'Depth Pro', background blurring and subject selection become faster and more accurate, allowing high-quality bokeh to be achieved even on smartphones. Since the AI understands depth information at the pixel level, effects that highlight fine hairs and leaves are also possible.

2024-10-apple-depth-pro-3d-image-9

Photo by littlekiss photography

Enhancing Reality in AR and VR

'Depth Pro' also significantly contributes to improving the quality of AR and VR. Unlike traditional depth recognition models, it can instantly estimate depth without camera metadata, accurately placing virtual objects in real space.

This enhances visualization of furniture placement and real engagement in games.

Applications in Autonomous Driving and Future Prospects

The speed and accuracy of 'Depth Pro' are ideal for depth detection in autonomous driving using onboard cameras. The AI enhances safety by recognizing obstacles on the road in real-time.

2024-10-apple-depth-pro-3d-image-14

Photo by Yuzurigima

Apple has already open-sourced this technology, and its use is expected to expand across more fields in the future.