[August 17, 2017] |
|
Visual Search Goes Open Source
Visual Search is now open source.
Zappen® (https://www.zappen.co)
and Yes Exactly, Inc.® today announced a FOSS production-ready framework
for mobile application development utilizing computer vision, image
recognition, and augmented reality (AR) technologies. Previously,
working implementations of visual search have only been available
through proprietary APIs such as that offered by Google (News - Alert) Lens (still in
beta).
"Zappen is like a QR Code on steroids," says Yes Exactly CEO Elizabeth
Gadwa. "You can take a picture with your phone and immediately connect
to websites, apps, and social media."
Discogs, the biggest and most comprehensive music database and
international marketplace, plans to integrate Zappen visual search
technology into its popular music database and collection management
app. "At the core of Discogs is our commitment to embrace and contribute
to the open source community," says Chad Dahlstrom, COO at Discogs. "Our
work with Zappen puts that commitment in action by contributing to
Zappen's ability to improve image recognition with scalability testing.
We're excited to b a part of this continuing innovation with the Zappen
team as we begin working to implement this technology into the Discogs
App."
Zappen combines a cross-platform React Native app with a PASTEC / ORB /
OpenCV backend employing the "visual words" algorithm for image
recognition and search.
"We were able to boost our image match success rate to over 95% by
optimizing hardware settings and server-side processing filters," says
Zappen cofounder Erik Amlee. "Having a team with a strong technical
graphics background was key."
Server-side image recognition confers several benefits over QR,
including enhanced security, brand recognition, and the ability to
re-use existing collateral and signage.
Zappen returns 98.7% accurate results in 3-5 seconds, under normal
lighting and bandwidth conditions. It has been tested for accuracy with
a dataset containing 100,000 images and is currently undergoing further
scalability and performance testing. The technology is best suited for
2D images such as logos, posters, signs, and product packaging.
Potential uses include public key verification, electronic currency
transactions, signs, exhibits, and advertising.
The Zappen visual search repository is available on Github under the
LGPL license at https://github.com/ZappenVDNS.
Future planned enhancements include machine learning and integration
with the EmerDNS distributed domain name system.
Try adding your own images and links (no coding required) with our free
demo at https://zappen.co.
View source version on businesswire.com: http://www.businesswire.com/news/home/20170817006094/en/
[ Back To Mobile World Congress's Homepage ]
|