Panasonic Releases API for Facial Recognition Utilizing Deep Learning Technology
Panasonic Corp. has started offering an application programming interface (API) for its facial recognition technology to the market in Japan.
Panasonic first began R&D into facial recognition in the early 1990s, and has been incorporating the technology into consumer products and in camera-based security solutions. The company has gone on to provide reliable and efficient facial recognition solutions such as for passport control at airports and ticketless entry at amusement parks.
Now, Panasonic brings this technology to more consumers. The facial recognition technology features a
quantity generation method of recognizing the whole face and certain parts of the face through the integration of multiple deep learning technologies. A proprietary Panasonic algorithm performs a degree of similarity calculation method to match faces and control errors caused by photographic conditions. Panasonic claims its facial recognition technology is able to accurately identify faces even under difficult conditions which previously presented challenges, such as when the face is seen from an angle, in difficult lighting conditions, or when the face is partially covered by sunglasses or a surgical mask.
Panasonic’s deep learning-based facial recognition technology can now be used by opening this API, making it possible to incorporate the technology into the company's customers’ smartphone applications, websites, and building access management systems. This enables entry/exit management at multiple business sites, streamlines movement within offices or work sites, and prevents identity theft. The facial recognition API is a micro service within Panasonic’s μSockets B2B IoT service which is currently available in Japan.
Panasonic is charging the facial recognition service on a monthly basis:
- Monthly service charge (excl. tax) = number of faces registered + number of facial recognitions
- Face registration charge: 5 yen/person
- Facial recognition charge: 1 yen/time
For example, in case 200 faces are registered, and facial recognition is performed a total of 16,000 times per month (4 times/person/day, 200 people, 20 business days) (5 yen/person x 200 people) + (1 yen/time x 4 times/person/day x 200 people x 20 business days), the fee is 17,000 yen/month.