Classifying DeepSat-6 Satellite Images Using Convolution Neural Networks

In this project, I’m using the DeepSat-6 satellite dataset which contains images extracted from the National Agriculture Imagery Program (NAIP) dataset to build a multi-class classification algorithm using convolution neural networks to assign a label for each image among several labels. The original dataset was extracted from Kaggle.

The following files and folders are included in this repository:

Final Report and Slides:

1. Final Report Report containing details on the data wranngling, exploratory data analysis, CNN models and summary.
2. Slide Deck Powerpoint presentation with an overview of the project and findings.

Google Colab Notebooks:

1. Loading Data
2. Exploratory Data Analysis
3. Random Forest Classifier
4. Baseline CNN Model
5. Transfer Learning 1 (with padding)
6. Transfer Learning 2 (with Upsampling)
7. Transfer Learning 3 (with fine tuning)
8. Model Comparison