"datasets for B-SGP"

"This submission contains two datasets for deep learning research on cloud imagery. The first dataset is in VOC format and is intended for segmentation model training, providing labeled images with pixel-wise annotations. The second dataset is in MNIST-like format and is designed for temporal sequence prediction model training, containing sequential image data suitable for developing and evaluating predictive models. Together, these datasets facilitate reproducible experiments in cloud segmentation and short-term cloud evolution forecasting."