Have you ever spent hours in line waiting for your turn to get your car repaired? Isn't it a pain in the neck?
One of the primary reasons for the delay is that everyone who visits wants an estimate of how much they might have to pay.
Our project aims to reduce this time.
Car Genesis analyses user's car for damages such as -
- Head Lamp
- Door Scratch
- Glass Shatter
- Tail Lamp
- Bumper Dent
- Door Dent
- Bumper Scratch
All the user has to do is choose where they live and upload some images of their car (min 1, max 4).
The ML model in the backend will analyse the car damages and the Android App / Web-App would show the estimated price and time for repair.
Just like when people put items in their shopping trolley at a wallmart store, people should be able to drive through the Garage's entrance and a screen or mobile app
should be able to give them estimated repair cost.
Alternatively, people should be able to calculate estimated repair cost all by themselves using our app.
There was huge imbalance in the dataset that we opted for. We tried to train the model again and again but the performance was unsatisfactory.
We tackled this problem by applying Image Augmentation on images with selected labels.
The model that we built from scratch was only able to produce an accuracy of around 60% so we performed Transfer Learning using MobileNetV2 instead for a better accuracy of around 80% and smaller size so that it could be directly deployed on edge devices.
We had a hard time connecting Android App with Fast API server as well that we tackled by replacing the package for POST request.
We feel accomplished to come up with such an innovative idea that can have real impact on people's lives and successfully implementing it in such a short frame of time.
- We learnt about the ways we could use to perform multi-label classification.
- This is the first time we collaborated with an APP developer which was a fun and learning experience as well.
- Improving the accuracy of the model
- Adding a tab where people can see quick-fix/DIY videos for above stated car defects
- Web App - Go to the backend folder and run
pip install -r requirements.txt
python manage.py runserver
- Android App - Download Car Genesis.apk and install it on your android device now.