Ridge-i, a company engaged in consulting and development work in the domains of AI and deep learning technology, launched the beta version of “DeepFire,” an AI surveillance service it has developed to detect abnormalities in facilities based on time-series data obtained from video footages, on March 25. On May 12, the company also announced that it will collaborate with Sompo Risk Management on raising awareness and utilization of the service upon the launch of its beta version, and begin offering it on a trial basis to companies in the energy, waste treatment, and manufacturing sectors.

Ridge-i is a technology company that serves as an integrated platform offering services ranging from assessment, development, implementation, to client-driven operations in the domains of AI and deep learning. Its track record hitherto has included the joint development of an automatic colorization technology for black-and-white video footages with NHK Art which has been used for NHK programs. It has also jointly developed a deep learning platform with Ebara Environmental Plant that can recognize different characteristics of waste products. This platform is currently being utilized in “waste-identification AI-equipped automatic cranes” at a waste incineration facility operated by Funabashi City. In recognition of its initiatives aimed at tackling social issues, the company received the METI Space Development and Utilization Award for its “Landslide Analysis Deep Learning Platform” (commissioned by JAXA).

The AI video surveillance service DeepFire whose beta version has been launched combines a deep learning model optimized for time-series analysis with advanced image processing technology. This allows DeepFire to accurately analyze states such as “combustion,” “viscous flows,” and “liquids,” whose complex transitions in their time series had made them difficult targets for quantitative analysis previously, and serve as a solution capable of automatically detecting abnormalities and symptoms suggestive of abnormalities.

By combining multiple networks that process time series data, DeepFire achieves a high inferential precision in advanced video analysis and learns from master data comprising data from normal settings as well as a small sample of data from abnormal settings. This allows the service to make determinations for any given setting with an accuracy on par with human experts by drawing on a small sample of learning data. Applications for a technology patent and trademark registration for this solution have been submitted in April.

When DeepFire is implemented for the surveillance of combustion sites, it can monitor the video footages captured by cameras installed in small thermal power plant boilers and combustion chambers of waste incineration facilities in real time. DeepFire is then able to detect symptoms suggestive of abnormalities from the video data of these combustion sites, which allows on-site personnel to perform the appropriate control and maintenance work before abnormalities arise, thereby enhancing the efficiency and rate of site operations.

As for the surveillance of processes involving viscous flows in food production, DeepFire ensures uniform product quality by allowing the state of fermentation and phase of kneading to be determined from images, something which is usually performed by experienced professionals via visual inspection. This also offers the advantage of making it easier for these professionals to pass on their skills.

By implementing DeepFire for the surveillance of liquid treatment operations, the determination of water quality in industrial sewage and wastewater treatment facilities as well as in water treatment facilities can be performed by constant AI surveillance instead of relying on manual visual inspection. This allows for a reduction in cost and makes the constant monitoring of water quality possible.


Provision of DeepFire on a trial basis to companies in the energy and waste treatment sectors

Ridge-i has also announced that it will collaborate with Sompo Risk Management on raising awareness and utilization of DeepFire upon the launch of its beta version, and begin implementing it on a trial basis to companies in the energy, waste treatment, and manufacturing sectors that are clients of Ridge-i's consulting service on the “implementation of AI video surveillance in production facilities”. DeepFire will be officially launched in the winter of 2020. Ridge-i is aiming to collaborate with Sompo Risk Management in the medium to long term on the utilization of the DeepFire model and various forms of data held by clients.