Expenses-OCR

Mobile App using OCR to process invoces | BSc project | 2018/2019 | Supervisor

The problem was introduced by an external company as a challenge (Softinsa/IBM) and independently supervised by me at IPT and developed by Hélder Lopes and Edgar Oliveira: a solution to improve how collaborators report travel expenses. The students developed an Android application (prototype), which uses optical characters recognition and machine learning, to scan invoices (e.g., meal, gas, train) and recognize the type and expense, introducing it in the system.

ExpensesOCR sequence diagram (in PT, from the BSc project report), and an invoice processed with OpenCV.

To this end, the students were tested several different solutions, settling with Android native development, OpenCV for OCR and MonkeyLearn to the the text classification into one of the possible categories.

Aplication screenshots for login, scanning the photo of an invoice and auto-filling the forms.
Application capturing an invoice and detecting the invoice edges.