S1-AAD: Sentinel-1 Amazon Airstrip Dataset

This repository contains the "S1-AAD: Sentinel-1 Amazon Airstrip Dataset", a dataset developed to support the training and evaluation of deep learning models for detecting, segmenting, and monitoring airstrips in the Brazilian Amazon Rainforest using Synthetic Aperture Radar (SAR) imagery.

The Brazilian Amazon Rainforest is a region of immense ecological importance, but it faces significant challenges from illegal human activities, such as the construction of unauthorized airstrips. Monitoring these activities is often hindered by persistent cloud cover, which limits the effectiveness of optical satellites. SAR imagery provides a crucial all-weather, day-and-night alternative for surveillance.

This dataset consists of 1,040 C-band SAR images from the Sentinel-1 satellite , corresponding to known airstrip locations sourced from the MapBiomas project. It is structured to support three distinct machine learning tasks:
1. Object Detection
2. Semantic Segmentation
3. Change Detection