Canon and Canon Marketing Japan announced on September 16th the "Dual Pixel 3D Converter," a Windows app that generates 3D models from photos. It will be available starting October 8th, initially for free.
The app can generate a 3D model from a single photo taken with a compatible EOS R series camera and lens. There's no need to capture numerous photos from multiple directions or prepare specialized 3D measurement equipment like LiDAR.
For 3D conversion, the app utilizes Canon's "Dual Pixel CMOS AF" autofocus technology. It works by leveraging information obtained from the two photodiodes in each pixel of the image sensor to add depth to the image.
It supports both the 3D reconstruction of an entire scene, including the background, and the generation of 3D models with the main subject isolated. A "Dual Pixel 3D SDK" for corporate clients is also planned for release within 2026, with anticipated uses including product displays on e-commerce sites and advertising.
While services that create 3D models from photos are becoming more common, Canon's approach stands out. Instead of merely estimating depth using generative AI, it leverages information that the camera originally acquired for autofocus. This can be described as an approach unique to a camera manufacturer.