INCYDE: A large scale cyclone detection and intensity estimation dataset using satellite infrared imagery
Keywords: Remote Sensing, Convolutional Neural Network, Disaster Management
Abstract: Tropical cyclones are devastating natural phenomena that cause a significant amount of damage every year. Conventionally, the Dvorak technique is used to detect cyclones and estimate cyclone intensity from satellite infrared imagery by observing cloud patterns. Satellite infrared imagery provides valuable information for detecting cyclonic storms. Recently, deep CNN models have proven to be highly efficient in detecting relevant patterns in the images. In this work, a novel cyclone detection and intensity estimation dataset called INCYDE (INSAT-based Cyclone Detection and Intensity Estimation) dataset is presented. The cyclone images in the dataset are captured from INSAT 3D/3DR satellites over the Indian Ocean. The proposed INCYDE dataset contains over 21k cyclone images taken from cyclones over the Indian Ocean from the year 2013 to 2021. The dataset pertains to two specific tasks: cyclone detection as an object detection task, and
intensity estimation as a regression task. In addition to the dataset, this study in troduces baseline models that were trained on the newly presented dataset. The results of this research would help develop innovative cyclone detection and intensity estimation models, which in turn could help save lives.
Primary Area: datasets and benchmarks
Code Of Ethics: I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics.
Submission Guidelines: I certify that this submission complies with the submission instructions as described on https://iclr.cc/Conferences/2024/AuthorGuide.
Anonymous Url: I certify that there is no URL (e.g., github page) that could be used to find authors' identity.
No Acknowledgement Section: I certify that there is no acknowledgement section in this submission for double blind review.
Submission Number: 7462
Loading