Overview / Usage

This project is based on the Intel Scene Classification Challenge which was released by Analytics Vidhya in collaboration with Intel.

The dataset consists of the different types of images (of 150x150 dimensions) which are nothing but natural scenes. The categories particularly were given as:

• Buildings
• Forest
• Glacier
• Mountain
• Sea
• Street

The task is to classify images with respect to the above categories.

Methodology / Approach

The dataset has a decent amount of similarity with the classic Imagenet dataset. So the choice of using a model already trained on the Imagenet dataset came quite naturally to me. The next challenge was to fine-tune the model and set the hyperparameters accordingly. I have detailed my approach in the accompanying notebook of the project. Readers are requested to take a look at it and let me know what they think.

Technologies Used

Python
Fastai
Google Cloud Platform

Repository

https://github.com/sayakpaul/Intel-Scene-Classification-challenge