![]() ![]() I advise you to be very careful in this step. The train directory can have about 80% of the images and the test directory can have the remaining 20%. Break the dataset into two directories – train and test. You can take such photos with your mobile phone or scrape from the internet. We can create a dataset on images with number plates like the ones above. The latter 3 will be covered in a later blog. We will cover the former 4 steps in this blog. Identify shortcomings and explore methods to improve the model.Read the number plate using an OCR tool.Extract the number plate using the trained model.Train an existing object detection model to detect number plates in a picture.Create a dataset of images with number plates.We will break down the task of building a custom number plate reader to the following LabelImg – LabelImg is a graphical image annotation, which we will use in labelling our datasets. ![]() Python -We will write all code in Python 3.It enables us to read text embedded in images. Pytesseract – Pytesseract is a python wrapper for Tesseract an Optical Character Recogniser tool.OpenCV has a bunch of tools to manage pictures, videos and algorithms that manipulate images. It is written in C++, we will be using its Python extensions. OpenCV – OpenCV or Open Computer Vision is the most popular tool for computer vision.You can find a list of all available models here. We will use the SSD inception v2 model as it gives us a good balance of both accuracy and speed. Tensorflow object detection API can use several models for object detection. SSD Inception v2 model – The SSD or single shot detector lets us detect and localise objects in an image with a single pass or a single shot. ![]() It lets us construct, train and deploy a variety of object detection models. This API is an open-source framework built on top of TensorFlow. ![]()
0 Comments
Leave a Reply. |
AuthorWrite something about yourself. No need to be fancy, just an overview. ArchivesCategories |