News Releases

TOKYO—Toshiba Corporation (TOKYO: 6502) today announced the market launch of GridDB Vector Edition, a high-speed database that applies vector analysis to big data sets. GridDB integrates the data analysis technology Toshiba announced in May, including the ability to find the image of a single individual in a database of 10 million photographs in only 8.31 milliseconds. Toshiba initially aims to propose the product as a facial recognition solution.

In mass facial recognition systems, like those deployed at immigration gates in ports of entry and borders, matches are achieved by vectoring^[1] facial data extracted from a photograph of an individual, and then running an analysis against the data set. Utilizing GridDB Vector Edition for this boost the search speed to 50 times^[2] that of the conventional technology.

Vector Indexing Technology is an original technology developed by Toshiba. It builds groups of similar vectors, which realizes fast identification of groups close to the vector in any given query. The need to compute the distance between individual vectors and that in the query is eliminated, realizing ultra-fast vector analysis.

High-dimension vector data is used in image matching and machine learning for big-data analysis and large-scale media analysis. However, data volumes continue to grow exponentially, and data verification for computing and recognition must keep up with this information explosion. GridDB Vector Edition ends the main drawback associated with using major data matching systems and its predecessors—the time it takes.

Toshiba will further enhance the application for this database, and promote its application to areas such as financial data mining systems.

Highlighted Features

1. Realization of ultra-high-speed vector processing by vector indexing
GridDB Vector Edition boost vector processing speed 50 times by expressing data in high-dimension vectors and pre-indexing similar vector groups.

2. Pattern analyze by expandable SQL (structured query language)^[3]
It can easily be applied to pattern analysis by applying pattern reference enhanced SQL.

3. Flexible expandability and high availability
The system can be scaled up without any need to stop the system, depending on user capacity and performance requirements. It is also highly resilient, able to continue operation even in the event of localized system failure.

Note:

 

[1]

Feature data expressed as high dimensional vectors, with 100-100,000 dimensions; many more dimensions than 2D (planar) and 3D (spatial) vectors.

[2]

	elapsed time (msec)	Precision of face recognition
This technology:	8.31	98% and over
Preceding technology A:	449.3
Preceding technology B:	389.5

For more information, please see the following press release.
“Toshiba's Ultra-Fast Data Matching Technology is Over 50 Times Faster than its Predecessors” https://www.global.toshiba/ww/news/corporate/2016/05/pr2501.html

[3]

Enhanced Structured Query Language (SQL): A database language for performing operations and data definition and control, including data insert, query, update and delete, that also provides a method for writing compounds statements, such as variable and conditional statements.

■ About GridDB:
GridDB is a scalable database developed by Toshiba for ultra-high-speed processing of big data in IoT and infrastructure. Data conventionally stored on hard disks is placed in the main memory of the server by in-memory architecture, to realize high-speed database processing at petabyte levels of big data. Expansion is easily achieved in line with data requirements and processing performance, with the advantage of no need to halt operation.