The AllRead Software is a deep learning-based information spotting technology able to read in a single shot the relevant structured alphanumeric text appearing in an image.
A convolutional neural network (CNN) architecture is trained in an end-to-end manner and is able to directly output readings without any explicit text localization step. By bypassing any segmentation step and automatically ignoring irrelevant textual content, we provide a 100% automatic, smart, end-to-end and high precision reading system.
01 / IN THE WILD
No need to explicitly localize or point to the relevant information, it will be automatically spotted.
02 / AGILE
Pioneering deep neural network model that can be rapidly trained and adapted to any application scenario.
03 / ACCURATE
Recognition models are ad hoc tailored to work in specific scenarios, guaranteeing the highest recognition accuracy.
04 / INTELLIGENT
Targeting exclusively important information for the user, ignoring the rest of non-relevant information.
05 / LIGHTWEIGHT
Can run directly on mobile devices, or on the server side communicating with the client through a web-service interface.
Using machine learning, AllRead MLT disrupts the traditional Optical Character Recognition (OCR) technology, which needs a lot of preliminary configuration and subsequent control. With OCR, the text must be correctly framed in the image to be read. Every character is segmented, multiplying the risk of errors. In other words, OCR technology is not robust, and therefore its output is not reliable.
AllRead goes way beyond OCR.
AllRead directly outputs readings without any explicit text localization step. It spots the relevant text in the image, wherever it is, and resists to movement, blurr, dust and deformation. AllRead MLT´s recognition models are rapidly adapted to new reading scenarios, and its constant learning curve reaches the highest accuracy, even in the most challenging environments.