12-01, 13:50–14:10 (Asia/Seoul), Circle Room
Flood Detection from Sentinel-1 imagery has traditionally relied on thresholding approaches. Conventional methods often face challenges in regions where the surrounding areas also exhibit low backscatter, resulting in inaccurate results. Furthermore, processing and conducting large-scale analysis pose significant challenges.
In this presentation, we aim to overcome these limitations by harnessing the power of Deep Neural Networks (DNN) and supplementary data, such as elevation and land-cover information, to enhance flood detection accuracy. This approach centers around training a neural network using open source cloud2street flood dataset. To achieve large scale analysis we leverage the capabilities of the Cloud Native Geospatial Tools, which helps us avoid downloading data when performing inference.
Aman is a Geospatial Scientist at GalaxEye Space with experience in Python and Deep Learning. He has a keen interest in open specification spatial data formats. A regular blogger, Aman has an active YouTube and GitHub handles, where he shares detailed examples of remote sensing and GIS case-studies. Find him at https://amanbagrecha.github.io/