LineOCR is a text detection and recognition system, based on our research in Artificial Intelligence.

Our goal is to create a system, that can read text from images as good as humans do. In other words, if you can read the text, our system should be able too.

<aside> 💪 What we have: a prototype, that demonstrates our system capabilities

</aside>

<aside> 🌚 What we need: rewrite the prototype to maximize system performance and quality

</aside>

Main features:

1. Can read handwritten text, digital born, real world and incidental scene text images
2. Self-learning (and generates explicit rules, more on that below)
3. Can deal with blurred, flashed, grainy, noisy etc. images

Can read handwritten text

One of the hardest OCR problems is reading continuous handwritten text. There are very few OCR engines, that even try to do that, and none of those have satisfactory results. We claim to have solved this problem.

Can read Digital Born images

Screenshots, images, created in Photoshop, memes etc. These images have specific problems, that sometimes need a special approach, or even a special OCR system altogether. LineOCR comes with this functionality included.