inkCapture
Handwrittent text recognition
Application for digitization of handwritten fonts
Printed documents (such as books and newspapers) are usually digitized with textual information and allow direct text searches within the documents. , currently there is no such possibility for direct text searches within handwritten texts (eg. chronicles, diaries, notes). As a result, researchers spend hours physically looking through hundreds of pages of documents in search of information. Also there are not enough accurate and reliable tools for the automatic recognition of handwritten text.
Digitization of manuscripts using artificial intelligence
The inkCapture application radically changes the possibilities of digitizing historical handwritten documents.
Unique applications for handwritten text recognition
inkCapture aims to change this. Deploying modern machine learning and computer vision methods will enable automated text extraction from handwritten documents and fundamentally enhance the searchability of historical documents.
Using artificial intelligence
InkCapture uses machine learning methods and neural networks to enable the system to recognize handwriting in different styles. The system has the ability to learn. Its success rate is largely influenced by the amount of data on which it can train letter and word recognition in advance.
Search your collections
You can search all your collections. You can search not only on the basis of the text you enter, but also on the basis of an image. A key feature of the entire solution is the continuous learning and improvement of skills based on feedback from text recognition and search.
Access your documents from anywhere
Responsive design adapts to the display and allows you to browse manuscripts anytime, anywhere. You can access collections from your mobile or tablet. A mobile app will also be available soon.
We are teaching neurons to know manuscripts
We use municipal chronicles, archival documents and papers, private historical documents, handwritten records and similar items to teach the neural network as much as possible. We collaborate with academic and historical institutions to develop solutions. We are currently in the development phase. A demo version will be available from January 2022 in Czech language.