Semantic segmentation of drone UAV images to improve disaster management by detecting and segmenting flood-affected areas. The model is trained on the Floodnet dataset.
This project aims to enhance disaster management by using a deep learning model to perform semantic segmentation on drone images. The model helps in identifying and segmenting flood-affected areas, providing valuable information for response and recovery efforts. It can classify the given image into 10 classes - ('Background':0, 'Building-flooded':1, 'Building-non-flooded':2, 'Road-flooded':3, 'Road-non-flooded':4, 'Water':5, 'Tree':6, 'Vehicle':7, 'Pool':8, 'Grass':9)
The Streamlit app provides an interface to upload images and view the segmentation results in real-time. Streamlit Demonstration - Click here
The model is trained on the Floodnet dataset. You can download the dataset from Floodnet Dataset. Please also refer to the research paper paper .
We used Iou per each class to measure the accuracy of our model.
This project is licensed under the MIT License - see the LICENSE file for details.
We would like to thank the developers of the Floodnet dataset and the contributors to open-source deep learning libraries that made this project possible.
Citations for papers referred - @ARTICLE{9460988, author={Rahnemoonfar, Maryam and Chowdhury, Tashnim and Sarkar, Argho and Varshney, Debvrat and Yari, Masoud and Murphy, Robin Roberson}, journal={IEEE Access}, title={FloodNet: A High Resolution Aerial Imagery Dataset for Post Flood Scene Understanding}, year={2021}, volume={9}, number={}, pages={89644-89654}, doi={10.1109/ACCESS.2021.3090981} }
@article{rahnemoonfar2020floodnet, title={FloodNet: A High Resolution Aerial Imagery Dataset for Post Flood Scene Understanding}, author={Rahnemoonfar, Maryam and Chowdhury, Tashnim and Sarkar, Argho and Varshney, Debvrat and Yari, Masoud and Murphy, Robin}, journal={arXiv preprint arXiv:2012.02951}, year={2020} }