Car-Brand-Classification

Deep-learning based web application that will predict the brand of a car.

What is this project about?

• I created this project to get started with deep learning. Specifically, how to predict classes using images via Deep Learning.
• I have used the concept of transfer learning. Transfer learning is a deep learning method, where a model developed previously for a task, is reused as the starting point for a model on a secondary task. More about transfer-learning, here.

Tech-stack used:

Algorithms used:

As mentioned earlier, I've used transfer learning in this project. Specifically, I've used ResNet 50 algorithm. More about ResNet50, here.

Dataset used:

Dataset is provided in the repository itself.

Accuracy score:

After running the model for 30 epochs, I was able to achieve:

• Accuracy: 87.50%
• Validation-accuracy: 70.69%;

Demo Video:

Issues:

• The UI of the web app could be improved.
• The accuracy can be increased by using custom neural nets, or by using algorithms like Inception V3, etc.