image recognition


Image recognition is a Computer Vision technology that makes it possible to analyze the resemblance between two pictures (or the objects pictured in them).


This technology is also the foundation of the mechanics of marker recognition used by Pikkart’s Augmented Reality apps.



Computer Vision can be used for:
Content-based image retrieval: image search based on visually similar images


Logo recognition: recognizing company logos in pictures and videos


Place recognition: using the camera of a mobile device to recognize a place



This technology can be applied to many fields:



Marketing and events: brand/logo recognition
Museums: image/artwork recognition


Advertising: flyer and poster recognition
Tourism: building and landmark recognition


Videogames: playing card recognition



Image recognition opens up new possibilities on mobile and industrial applications.


Algorithms being able to automatically recognize objects and places creates new opportunities for the creation of Intelligent Applications that can entertain, inform, and help users through natural interactions.


Pikkart’s image recognition system is used in our Cloud Recognition System (CRS) to perform fast image searchs in a dataset of hundreds of thousands of pictures. When a user frames a marker with their camera, the image is sent to cloud servers that, using advanced Computer Vision algorithms, find the matching image (if it exists) in less than one second.

Pikkart CRS can also be integrated in existing server installations in order to have a customized and independent service of image recognition.

When is cloud image recognition useful?

When your application requires a dynamic dataset: the cloud image recognition system can be configured via our CMS to instantly change which markers are available for recognition, thus changing the behavior of the application on the client side in a transparent way for the final user. 
When you have huge databases of images: the cloud image recognition system is scalable to hundreds of thousands of images, removing the limitations of mobile devices in terms of memory and computational power.  



Relevant publications

“Segmentation models diversity for object proposals”, Manfredi et al., Computer Vision and Image Understanding, 2016


“GOLD: Gaussians of Local Descriptors for image representation”, Serra et al., Computer Vision and Image Understanding, 2015





Last update:  22/03/2018