HANDS-ON-LAB

Water Body Identification from Satellite Images Project

Problem Statement

A collection of water bodies images captured by the Sentinel-2 Satellite. Each image comes with a black and white mask where white represents water, and black represents something else but water. The masks were generated by calculating the NWDI (Normalized Water Difference Index), which is frequently used to detect and measure vegetation in satellite images, but a greater threshold was used to detect water bodies.

Dataset

Kindly download the data from here.

Tasks

  1. Create the config.yaml file with all the appropriate details as given in the medical image segmentation project.

  2. Create the DataSet loader class that is used to get the image, its mask and the image_id for each of the training/validation/test images.

  3. Create the network.py,  utils.py, train.py, predict.py, validate.py file similar to the one discussed in the project

  4. Create the engine.py file and run the model for this water body identification project.

  5. Analyze the results on the test data to see how good the model is able to segment the images.

Analyze the results of the water body identification model on the test data for accurate segmentation.

 

FAQs

Q1. How were the water masks generated?

The water masks were generated by calculating the NWDI (Normalized Water Difference Index) with a higher threshold to detect water bodies.

Q2. What is the purpose of the config.yaml file?

The config.yaml file contains project-specific details, such as file paths, hyperparameters, and other configurations for the water body identification project.

Q3. How can I assess the model's performance?

You can analyze the results on the test data to evaluate how well the model is able to accurately segment the water body images.