Business cards detection & OCR
General purpose
Overview
Compiling virtual databases of business cards for an online verification firm. Business cards come in many forms & formats & nowadays taking your business online means opening a huge market for your product. And lots of business owners daily apply for verification & getting information from images manually is time consuming. For that we were tasked to implement a deep learning model which can be used for any background & any color of card to extract information from it.
Technology
- Deep Learning
- Neural Networks
- Computer Vision
- Image Processing
Technology
- Python
Key Technical Challenges:
- Background for extracting usually must be the same but here it was not so. We had to work our way in building such models.
- Different fonts for possibly every business card we come across. Had to work to make the model more generalized.
Business + Technical Points:
- Effective detection for every card’s api fired
- Possibility for reselling the detection API in place
- Should work with low-resolution images with more than 50% detection ratio.
Result:
- Robust model with resellable API with standard security protocols.
We use cookies to improve your browsing experience. Learn about our 