Hng Task 1 Plate Number Detector

Published on Sep 20th, 2019 06:41 AM

Hi there!!!

I am so glad to be part of this intership. Within these few days another side of me I never knew exist was activated, and I must say I am very impressed. Thank you Hotel.Ng for this opportunity to find myself.

The tasked was to collect 50 images of Plate Numbers which most of it should be Nigeria plate numbers and 50 images of Non Plate Numbers and put them in different folders with name Plate Numbers and Negative Images respectfully. These forms a dataset that should be trained using machine learning, which when untrained dataset is passed through the model it will accurately predict whether it is a Plate Number Image or a Negative Image.

Note this is the first time I am getting to know about machine learning, so if your question right now is, was I confused on how to start? Yes!!! you guessed right. Thanks to internet, some links were shared and I started reading about machine learning especially about classification, Convolutional Neural Network, etc. I got more confused. I did not understand anything, I thought of changing track but I still want to do this machine learn. So I decided to follow up the professionals that have done theirs with 2 days. You know the saying, "Code along with the steps even when you don't understand the function, soon it will start making sence." I think I heard that from a programmer.

So I followed the steps, dived into the internet and got my datasets and put them in the appropriate folders. I tried several IDE but the one that worked for me is Colab notebook, because all the libraries that I needed are in colab.I defined a function that can read the image into a dataframe and imported an opencv python library that could be used to in reading the images and loading them into the dataframe. I ran into several bugs, thanks to fellow interns particularly CHIDI BEDE who really helped me tp debug the program. I could not have successfully done this without them. I resize the image to avoid being too large and reshaped it into dimensional array to make it easy to train using the scikit learn models.

Below are the screen shots