Datasets and Publications

General description of the dataset

This dataset comprises flood-related images collected from several publicly available datasets, as well as frames extracted and annotated from related videos. The training and validation sets were constructed using the following sources: “Flood Area Segmentation,” “Water Dataset,” and “Roadway Flooding Image Dataset.” For the first two sources, the corresponding binary masks were normalized to values in {0,1}, while the third source was retained in its original normalized form. The test set consists of video frames extracted from real flooding scenarios in Greece and Italy, which were annotated for segmentation purposes. Specifically, the Greek and Italian videos contributed 567 and 1,406 frames, respectively. If one uses any part of these datasets in his/her work, he/she is kindly asked to cite the following papers:

• P. Mentesidis, V. Mygdalis and I.Pitas, "Improve Real-time flood segmentation by encoding and distilling foreground information", IEEE International Conference on Image Processing (ICIP), Anchorage, Alaska, USA, 13-17 September, 2025.
• A. Gerontopoulos, D. Papaioannou, C. Papaioannidis and I.Pitas, "Real-Time Flood Water Segmentation with Deep Neural Networks", IEEE 25th International Symposium on Cluster, Cloud and Internet Computing Workshops (CCGridW), Tromsø, Norway, pp. 85-91, 2025

Dataset Structure

The dataset is organized into separate folders for the training, validation, and test sets, each containing the corresponding annotations and a CSV file specifying the image paths, annotation paths, and the source of each image. The training–validation split was performed using a 3:1 ratio. Finally, the URLs of the original data sources are provided in the “sources.csv” file.

Details on acquiring the dataset can be found here. 

The dataset will be used for wildfire image classification and burnt area segmentation tasks for Unmanned Aerial Vehicles. It is comprised of 5,408 frames of aerial views taken from 56 videos and 2 public datasets. From the D-Fire public dataset, 829 photographs were used; and from the Burned Area UAV public dataset 34 images were used. For the classification task, there are 5 classes (‘Burnt’, ‘Half-Burnt’, ’Non-Burnt’, ‘Fire’, ‘Smoke’). As for the segmentation task, 404 segmentation masks on a subset have been created, which assign to each pixel of the image the class ‘burnt’ or the class ‘non-burnt’. If one uses any part of these datasets in his/her work, he/she is kindly asked to cite the following paper:

• M. Siavrakas, C. Papaioannidis and I.Pitas, “BLAZE: A dataset for wildfire and burnt area UAV image classification and segmentation”, IEEE International Conference on Image Processing (ICIP), Anchorage, Alaska, USA, 13-17 September, 2025.

CSV files are provided containing the frames taken from every video, the class that has been assigned to them, the path to the respective segmentation mask along with the mask for the segmentation subset and the related links to the public videos and the 2 public datasets.More details on the dataset are available in the following papers:

• de Venâncio, P.V.A.B., Lisboa, A.C. & Barbosa, A.V. An automatic fire detection system based on deep convolutional neural networks for low-power, resource-constrained devices. Neural Comput & Applic 34, 15349–15368 (2022). DOI
• Tiago F.R. Ribeiro, Fernando Silva, José Moreira, Rogério Luís de C. Costa,Burned area semantic segmentation: A novel dataset and evaluation using convolutional networks,ISPRS Journal of Photogrammetry and Remote Sensing,Volume 202,2023,Pages 565-580,ISSN 0924-2716. DOI

Details on acquiring the dataset can be found here. 

General description of the dataset

The dataset comprises of 766 social media posts in Greek from the platform “Mastodon” spanning the 2023 wildfires in Greece. Each post was annotated internally with Plutchik-8 emotions. To obtain texts use the Mastodon API (https://docs.joinmastodon.org/api/) along with the provided IDs.

Dataset Structure

The dataset comprises a single CSV file containing pairs of text identifiers and their corresponding sentiment labels.

Details on acquiring the dataset can be found here.