AllRead innovating Technology



AllRead Software relies on unique deep learning-based recognition models providing highly robust reading in unconstrained environments “in the wild” – where traditional OCR tend to fail.

The solution is based on a Convolutional Neural Network (CNN) architecture trained in an end-to-end manner which can directly output readings without any explicit text localization or segmentation step. Bypassing the above steps, and automatically ignoring irrelevant textual content, AllRead yields a 100% automatic, smart, end-to-end and high precision reading system.

The AllRead technology is the result of an innovation (various scientific papers published) covering 5 years of research of three computer vision PhD researchers from the Computer Vision Center (CVC), Autonomous University of Barcelona.


Disrupting OCR Software and infrastructure



AllRead is robust to dust, dirt, movement, blur, partial occlusion, damage or rotation and when benchmarked with the latest Deep Learning-based reading approaches published from the academy, AllRead technology surpasses the SoA in all these challenge categories.

Unlike other tracking solutions, AllRead MLT’s solution is focused on the software, leaning on the benefits of Deep Learning rather than on hardware devices.

Hardware is fast becoming a commodity, which makes AllRead’s hardware-agnostic aspect one of its key features.

Furthermore, traditional OCRs often require that the camera be activated to initiate a reading, and that the text be framed perfectly in the image to be read. AllRead spots the text when it appears, wherever it is in the image, ignoring any potential text pollution. Consequently, the hardware needs (multiple cameras, sensors, lasers…) are much lower.