The project trains a CNN model to locate and read vehicle license plates from input images. It generates synthetic training data by compositing number plate images onto background photos, then uses TensorFlow for model training and inference. OpenCV is used for image preprocessing and post-processing tasks.