This work illustrates a design and development of an efficient Automatic number plate recognition system (ANPR), using image processing and deep learning techniques. The work focuses on the deep learning SSD MobNet model, which will be used to detect the number plate of the vehicle. We utilized two different sequential convolutional neural networks that are smart enough to recognize plates characters. Moreover, it is shown that with variations of pretrained architectures, one can obtain the desired trade-off between complexity and accuracy, and SSD MobeNet v2 and a simple CNN model can obtain the desired results with satisfying cost and accuracy.
In this work, we utilized two publicly available datasets from Kaggle.
All installation instructions for libraries are in the code notebook. You can use Jupyter Notebook.
This project was inspired by RealTime Object Detection and AI based indian license plate detection