Advancement of computer vision algorithms, remote sensing, and geospatial technologies allow a continuous monitoring, mapping, and analysis of feature events on the earth. The goal of this project was to build machine learning and computer vision models to help automate the process of classification and ships identification in multispectral satellite images. In this project, I processed and analyzed high-resolution raster dataset, built convolutional neural network (CNN) models (Grayscale input and RGB input) and deployed it in AWS.
TP (True Positive), FP (False Positive), FN (False Negative)
This project has used openly licensed Planet satellite images collected over the San Francisco Bay area, and provided by a Kaggle user::
- Planet Labs, Inc. (Earth imaging company): https://www.planet.com/products/open-california/
- Provided by a Kaggle user: rhammel @ https://www.kaggle.com/rhammell/ships-in-satellite-imagery/data