WineAdvisor is a wine social network. On this application, members can identify a bottle, save it and share it. Once the bottle is identified, the application returns a lot of information about the origin of the wine and its tasting. However, manually entering all the information on the label was tedious. Wineadvisor wanted to find a solution to quickly identify a bottle of wine.
Neovision developed an image recognition technology to accurately identify a bottle. To do this, simply take a picture of the bottle label via the application. This photo is then compared with more than a million references and the application retrieves all the information about the photographed bottle.
With this technology, WineAdvisor improves the UX of its application. Indeed, manual data entry generates a strong technological friction and remains a factor of error (type in wrong information). The use of visual search allows to limit this friction, to make the returned results more reliable and to place the Wineadvisor application in the growing technological paradigm of visual search.
Neovision designed and developed image recognition technology for the start-up company WineAdvisor.
This technology allows its mobile application to find the reference of a bottle of wine from a photograph of its label. It currently operates on a basis of 1,000,000 images and responds in less than 0.5 seconds.
The technology is resistant to changes in brightness, angle of view, occlusions and partial alterations of the label. We have also integrated an administration system that allows our customer to add, delete and modify the images in the database in real time.
Used on virtual machines, the technology is scalable thanks to Google Compute Engine load balancing and autoscaling.
At the same time, Neovision has also developed complementary Optical Character Recognition technology.
This OCR technology enables WineAdvisor’s mobile app to recognize the vintage of the bottle. Indeed, two bottles coming from the same domain can vary according to the year of production. A 2004 vintage is different from a 2010 vintage. By identifying the vintage, Wineadvisor brings back precise information to the user, such as the right price for example.
The algorithm has been designed and optimized to detect dates and works even on poor quality images
Today, Wineadvisor provides its users with a mature and efficient identification technology. And if you want to get your own idea about this technology, you can download the application right here:
Google Play : https://play.google.com/store/apps/details?id=com.wineadvisor
App Store : https://itunes.apple.com/fr/app/wine-advisor-vin-champagne/id944699586
COMPUTER VISION, MACHINE LEARNING, OCR, CLOUD COMPUTING
« Neovision worked on WineAdvisor’s image recognition tool. Their team was able to adapt to the surrounding context and meet our more than demanding requirements in terms of deadlines and efficiency. They totally met our needs and created interfaces and processes that met our expectations at the beginning of the project. Their knowledge of the state of the art algorithms and their understanding and programming skills enabled us to successfully complete our projects and I will not hesitate to call on Neovision again. »
Cédric Morelle, CTO and co-founder of WineAdvisor
16 October 2020
Application Mobile, Cloud, Computer Vision, Machine learning, Retail