8-Bit Quantized 38-Cloud: Cloud Segmentation with 5-bands

For our model training and processing, we utilized a modified version of the 38-Cloud: Cloud Segmentation in Satellite Images. To optimize memory allocation and computational efficiency, the raw imagery was quantized to an 8-bit representation. Additionally, the dataset was structured to include five distinct spectral channels: the native Red, Green, Blue (RGB), and Near-Infrared (NIR) bands, alongside an explicitly derived RGB Mean Intensity band.