vertify/forest-forward-biomass-dataset-goa

这个数据集包含了专为训练和验证深度学习模型而精选的多传感器卫星图像和相关的生物量参考数据，用于估算南亚地区（特别是印度的森林区域）的地上生物量（AGB）。数据来源于Sentinel-1、Sentinel-2、Landsat 8和PALSAR等卫星传感器，以及数字高程模型和林冠高度等辅助数据。参考生物量数据通过航空LiDAR获取，并通过广泛的现场采样进行校准。数据集具有40米的空分辨率和季节性的时间分辨率，共有59个输入波段。数据集包括卫星图像块、对应的地面真实生物量值和有效数据像素的二进制掩码。

This dataset contains multi-sensor satellite imagery and associated biomass reference data specifically curated for training and validating deep learning models, such as the Attention-based BiomassCNN, for estimating aboveground biomass (AGB) in Southern Asia. The data sources include satellite sensors like Sentinel-1, Sentinel-2, Landsat 8, and PALSAR, as well as ancillary data such as Digital Elevation Model (DEM) and Canopy Height. The reference biomass data are derived from aerial LiDAR and calibrated using extensive field sampling. The dataset has a spatial resolution of 40 meters, a temporal resolution of seasonal (three seasons in 2020), and a total of 59 input bands. It includes satellite image chips, corresponding ground-truth biomass values (Mg/ha), and binary masks indicating valid data pixels.