Refined Deep-SAR Oil Spill (SOS) dataset

This is the improved version of the Deep-SAR Oil Spill (SOS) datase [1]. To enhance training data quality, segmentation masks with significant annotation errors were manually corrected. Approximately 38% of the training masks and 50% of the validation masks were revised based on visual inspection and annotation consistency. These refinements aimed to reduce label noise and improve model performance during training and evaluation.

[1] Q. Zhu, Y. Zhang, Z. Li, X. Yan, Q. Guan, Y. Zhong,  L. Zhang, and D. Li, “Oil Spill Contextual and Boundary-Supervised Detection Network Based on Marine SAR Images,” IEEE Transactions on Geoscience and Remote Sensing, 2021.DOI: 10.1109/TGRS.2021.3115492