- Anomaly detection
- Sensor data recognition
- Status estimation
- State change detection
- Media recognition
Distance estimation with a monocular camera (aberration map 3D measurement)
Using a single photograph, measures distance to a subject and subject size in 0.3 seconds or less.
- Through supervised learning of the relationship between the information of blur caused by actual lens aberration, image position and distance to the subject, it became possible to quickly (*1) and accurately (*2) estimate the distance to the subject and the size of the subject from a single image.
- When processed using a graphics processing unit (Geforce GTX 1080Ti), with a neural network that applies an original acceleration method
- “Accurate”: Measurement error of approximately 10%; the same error as “stereo” method
Applications
- Infrastructure inspection service
- Automatic driving
- Product inspections
Benchmarks, strengths, and track record
- Original Toshiba technology that measures actual distance and actual size from a single photograph
- Can also measure distances of moving objects.
- Works with a variety of cameras, including smart phones, USB cameras, and single lens reflex cameras.
Inquiries
Please include the title “Toshiba AI Technology Catalog: Monocular depth estimation (aberration map 3D measurement)” or the URL in the inquiry text.
Please note that because this technology is currently the subject of R&D activities, immediate responses to inquiries may not be possible.
References:
- Press release, October 25, 2019: “Developed 3-D recognition AI that measures distance with the accuracy of a stereo camera using a commercial single lens camera”
- Press release, November 22, 2021: “Toshiba Develops World-first AI Capable of 3D Measurements using Only Photos from a Monocular Camera Zoom Lens”
- Toshiba Review, Vol. 75, No. 3 (May 2020); “An accurate distance measurement method based on aberration information from a single image taken with a single lens camera” (PDF)
- Toshiba clip (technology): “Will a single camera be the savior of infrastructure inspection? AI that supports the front lines, completed after multiple modifications” (in Japanese)
- Toshiba clip (episode): “See the future through a lens! Different knowledge, creating new technologies together” (in Japanese)
- M. Kashiwagi et al., “Deep Depth From Aberration Map,” ICCV, 2019.
- N. Mishima et al., “Physical Cue based Depth-Sensing by Color Coding with Deaberration Network,” BMVC, 2019. (PDF)
- M. Kashiwagi et al., “Monocular Depth Measurement with Lens Aberration Using a Smartphone Camera,” SSII, 2021. (in Japanese)
- N. Mishima et al., “Absolute Scale from Varifocal Monocular Camera through SfM and Defocus Combined,” BMVC, 2021. (PDF)
