Reading ingredients from a food product label

[Sabin Flo]

Hello everyone! I have this idea to develop an application that could tell people which ingredients are toxic in a supermarket product that they buy. The application would take a picture of the "ingredients" section of the label, would extract words and confront those words with a database of ingredients. I have tried tesseract on some pictures of labels( that were pretty clean - right focus, one color as a background ),  but the results were disastrous. Can I use tesseract on curved or rotated text? And if not do I have alternatives in the open source world that would perform good in this situations( products labels )?