Detecting and Classifying vehicles

Demo of my algorithm on image. (More details on my google-colab link)

Inspiration

My Inspiration to start this project is Google's Object Detection API and Tesla Auto driving. I mean what a great work they have done and now world is also moving towards to automation side So, I am also trying to solve public-parking security issue I faced. and In future this project will also help to build software for auto driving car.

What it does

here is some steps how my detection algorithm works

Firstly It detects a car in the frame.
after detecting a car it tries to find a car logo and license-plate.
it also finds any special logo or stickers like Uber, Ola, Taxi etc. because by using this stickers on a car I classify the vehicle type. -After detecting all things I applied some computer vision functionality and find the license-plate color ratio because it helps us to classify vehicle type.

How we built it

First of all I have to choose the best algorithm for object detection. So, I have done some research and find the deep learning algorithm called YOLO(You Only Look Once). at the begin I don't know about this then I read some Blogs and explore about YOLO then I have to find the car images dataset for training my algorithm after finding all images I have to make annotations on images to train our model. after annotation I have to update yolo configuration file according to our requirement and image dataset size. and also have to prepared datafile, configuration file and names file for yolo detection algorithm because I have used customized yolo.

Challenges we ran into

First of all biggest challenge is to find the images because I need proper car image with including all types(EV , Taxi) and with proper visible number plate for this project. Second challenge is to find the coordinates of detected part from image and store it for future process it's also little-bit challenging. Third challenging part is to classify car type using license plate because after detecting license place I have to find color ration it's also difficult little-bit.

These three are the most challenging part during my project.

Accomplishments that we're proud of

I scared about algorithm worked after training because I customized configuration file by self because I applied customized yolo and spend at least 15-16 hour only for training my images. during that training time, I thought I am wasting my time. but at the end it's worth and working good:)

What we learned

I learned some new technologies and explore new ideas and I also made some mistakes during this project. So , next time I will try to not repeat that same mistakes again. and also learned how manage and complete work in short time period.

What's next for Detecting and Classifying vehicles

-Next plan is to build a proper platform for this project and make it open source. -Also plan to get details of that particular car owner from license plat. -I also have other plan to make it capable for running in automatic car driving. It's difficult I know but will try . but for that I have to find a good team because It's very difficult to work solo for this project.