Extracting Structured Information from Invoice Images

Around 18 billion invoices are issued each year in USA and Europe alone. Form-like documents such as invoices, purchase orders, tax forms and insurance quotes are common in everyday business, but current techniques for processing these still employ a large amount of manual effort and time, or use OCR based heuristics for extraction. Although OCR has been fairly successful in helping digitisation of machine-printed text, the cost for such services are can multiply exponentially. By building a program to automate such an application would tremendously decrease the amount of money and manpower spent in physically processing these invoices making them more accessible and beneficial to the overall business and demand for these products. Furthermore, these automated systems will tremendously reduce the manual or human errors that creep in.