Skip to content

MustfizurRahaman/Effects-of-Label-Noise-on-Performance-of-Remote-Sensing-and-Deep-Learning-Based-Water-Body-Segmentat

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

20 Commits
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Basic overview

This is the repository for the paper "Effects of label noise on performance of remote sensing and deep learning-based water body segmentation models" . This repository contains:

Dataset

The dataset contains total 110 georeferenced clipped tiff files of Dhaka city collected using the GEE and also correspondent binary labels. They have been used for training, validating and Testing the model.

Notebook

The notebook contains the original model which have several sections containing data loader, data visualization, data augmentation, loss function, u-net model with encoder weights, and test function.

Requirements for the model

  1. pytorch
  2. skimage
  3. numpy
  4. opencv
  5. PIL
  6. segmentation_models_pytorch

Getting the repository

This repository can be downloaded and also cloned by the following steps, you can try either. git repository:

git clone https://github.com/monsurhillas007/LabelNoisePerformanceonWaterBodySegmentation.git

or download a zip archive.

Getting started with the code

After downloading or cloning the repository seperate your downloaded dataset in three individual folders such as Train,Valid,Test. seperate both image files and label files.

Before we run the code we create an 'anaconda' environment and install all the requirments on that virtual environment.

To create the environment:

conda create --name YOURENVNAME

After we create the virtual environment we must activate it:

conda activate YOURENVNAME

The code given in the repository is a notebook file which contains multiple cells working as individual python fie. to open the notebook type this in your anaconda prompt:

jupyter notebook

When jupyter notebook is open, adjust the necessary paths according to your needs and run the notbook.

Data augmentation

In this notebok, the labels of the dataset can be augmented in various ways. Different kind of augmentation functions are commented there. we can use those augmentation according to our need.

About

No description, website, or topics provided.

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published