As part of my time as Data Scientist at Shift Technology, I researched & implemented a license plate detection pipeline. The final solution consisted of a small web-server with an API, which could be used to send images and would return the position of car(s) and their respective license plate(s), and was used to detect fraud connected to car insurances.
My work was mainly to do a literature review of the various state-of-the-art computer vision algorithm (e.g. Faster-RCNN, YOLO). Then, I had to implement them starting from available open source solution, do transfer learning to detect license plates, and then build the server and client sides infrastructure.
Disclaimer: the image used for the post is not directly from my work, since the latter is property of the company. However, the result would’ve been extremely similar.