- Status estimation
- Media recognition
- Media data analysis
Human recognition (face recognition, action recognition, person re-identification between cameras, crowd density estimation)
Recognizes humans, from individual faces to groups, using the technology best suited to the distance from the camera.
- Demonstrates robust operation in various environments; e.g., camera installation angle and external light.
- Used in a wide range of fields, including security, marketing, and visual presentation of work processes.
- Click the “Play” button to play a video on YouTube.
- YouTube is a service unrelated to Toshiba Corp. Use of this service is subject to relevant terms and conditions.
Applications
- Security, wide area monitoring
- Marketing
- Settlements
- Visual presentation of work processes
Benchmarks, strengths, and track record
- No. 1 in Japan for the face recognition benchmark NIST FRVT 1:1 Verification (No. 9 in the world)
- VISABORDER Photos category; as of Oct 2021
- Institute of Image Information and Television Engineers, 47th (2019) Technology Promotion Award, Progress Development Award (Front-line Operation Division)
- Motion Picture and Television Engineering Society of Japan Inc., 73rd (2019) Technology Development Award
- Used in Toshiba Digital Solutions Corporation facial recognition AI “Kaometa.” (in Japanese)
Inquiries
Please include the title “Toshiba AI Technology Catalog: Human recognition (face recognition, action recognition, person re-identification between cameras, crowd densitiy estimation)” 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:
- V. Pham, et al., “COUNT Forest: CO-voting Uncertain Number of Targets using Random Forest for Crowd Density Estimation,” Proc. ICCV2015, pp.3253-3261, 2015.
- Shibata et.al.; “A system for calculating the trajectory of human movement between multiple cameras using Person Re-identification”; MIRU Symposium (on Image Recognition and Understanding), 2017. (in Japanese)
- Y. Shirakawa, et al., “Skip-Pose Vectors: Pose-based motion embedding using Encoder-Decoder models,” Proc. MVA, 2019
- Developed an AI technology to simultaneously and accurately monitor the movement paths of multiple persons based on images from multiple cameras
- Developed an AI technology to rapidly count the number of persons in group using a regular PC
- Yamaji, Shibata; “Method for learning group density estimates resistant to error in teaching data”; SSII2020.
- Kobayashi, Shibata; “Simultaneous estimate of multiple object tracking and object detection using Spatiotemporal Attention, MIRU2020.
- N. Kawamura and S. Kubota, “Sample-Dependent Distance for 1 : N Identification Via Discriminative Feature Selection”, ICPR2020.
- Toshiba Clip | Instantaneous density measurement from images: AI transcends the human eye (toshiba-clip.com) (in Japanese)
- Toshiba Clip | With a view toward society, moving forward together: Thoughts on the development of AI for measuring density (toshiba-clip.com) (in Japanese)
- https://www.global.toshiba/jp/technology/corporate/rdc/rd/topics/21/2111-02.html (in Japanese)
- The Toshiba Co-creation Center “Creative Circuit” Implements Video Recognition AI to Promote Value Creation with Co-creation Partners -Accelerating the social implementation of AI technology by sharing images using Toshiba’s latest video recognition AI- | Corporate Research & Development Center | Toshiba (global.toshiba)