Project for Vision and Cognitive Services exam. Solve a Kaggle challenge related to a Vision problem.
Link to Competition: https://www.kaggle.com/competitions/understanding_cloud_organization/overview
Link to Dataset: https://www.kaggle.com/competitions/understanding_cloud_organization/data
In this competition you will be identifying regions in satellite images that contain certain cloud formations, with label names: Fish, Flower, Gravel, Sugar. For each image in the test set, you must segment the regions of each cloud formation label. Each image has at least one cloud formation, and can possibly contain up to all all four.
The images were downloaded from NASA Worldview.
Shallow clouds play a huge role in determining the Earth’s climate. They are also difficult to understand and to represent in climate models. Part of the reason is that shallow clouds are not just the result of the global circulation of the atmosphere. Rather, they have a life of their own and arrange themselves in a variety of patterns. For many of these patterns, the basic mechanisms behind them are poorly understood. After some discussion, scientists agreed that there are four common patterns and called them Sugar, Flower, Fish and Gravel. By classifying these types of cloud organization, researchers hope to improve our physical understanding of these clouds, which in turn will help us build better climate models, in order to have better prediction of climate change or forecasting weather update
The aim of this project is to identify regions in satellite images that contain certain cloud formations, with names: Fish, Flower, Gravel, Sugar. For each image in the test set, we need to segment the regions of each cloud formation. This is a multiclass segmentation task where we have to find 4 different cloud patterns in the images.
The project is divided into two main parts. In the first one, I implemented a U-Net architecture from scratch and tried to find the best combination of hyperparameters to get the best results. Instead, in the second part I compare the U-Net architecture built from scratch with more complicated and already implemented networks, in order to see if they can achieve better results.