HLS-GPT: A Generative Pretrained Transformer (GPT) for Continental-Scale NASA Harmonized Landsat and Sentinel-2 (HLS) Reflectance Reconstruction Across All Bands on Arbitrary Dates
Researchers have developed HLS-GPT, a large-scale generative pretrained Transformer model designed to reconstruct NASA's Harmonized Landsat and Sentinel-2 (HLS) surface reflectance data. This model utilizes a hierarchical Transformer architecture to process varying spectral band configurations and operates on single-pixel time series. Trained on extensive data from the conterminous United States, HLS-GPT demonstrates robust reconstruction capabilities across diverse land surface conditions and outperforms conventional methods and the NASA-IBM Prithvi model in evaluations. AI
IMPACT This model advances AI's capability in processing and reconstructing complex satellite imagery for environmental monitoring.