Satellite-Image-Classification

Here is my first project based on Convolutional Neural Network (CNN) using Keras. It is a naive classifier which I used for my EVS Mini Project, classifying satellite images into three classes: agricultural, forest and river. I used UC Merced Dataset (http://weegee.vision.ucmerced.edu/datasets/landuse.html) for this project with 15 training images for each class and 3 test images for each class. Without much tuning, I was able to achieve 83% accuracy. My next approach will be to introduce class detection and segmentation using algorithms like Mask RCNN.