Role: Computer Vision Engineer
Today’s Menu is a web, mobile and chatbot application that enables users to discover, book and pay for the daily changing menu offered by restaurants. It aggregates and digitizes the daily changing menu offered by restaurants in real-time. It also provides allergen information for each dish on the menu. The way it is done is the manager of the Restaurant clicks the picture of the handwritten menu and uploads it to TMA's App. The rest is taken care of by the app.
My job in this project was to extract the text in digital form from the images. The images usually consisted of boards on which Menu was handwritten, so the main challenging part was to make an application efficient enough so that the menu can be converted properly. I used many image preprocessing techniques to improve image quality and then to find the area where the menu is written. Once the region of interest is extracted, OCR( optical character recognition) is performed on it. This gives digital text as an output that is utilized by the backend team to generate Menu in the App. To improve the accuracy of the result I also used services like Google Vision and Microsoft's API.