SSL4SAR
收藏资源简介:
SSL4SAR是一个未标记的SAR数据集,包含9,562张Sentinel-1图像和14张Sentinel-2图像,这些图像展示了北极14个不同大小和崩解前缘几何形状的冰川。数据集还包括每个冰川一张无云的光学图像,这些图像是在2020年6月至8月之间由Sentinel-2捕获的。SSL4SAR数据集是为了解决冰川崩解前缘提取问题而创建的,旨在为后续分析提供大时空数据库的崩解前缘位置。SSL4SAR数据集的创建是为了解决冰川崩解前缘提取问题,旨在为后续分析提供大时空数据库的崩解前缘位置。SSL4SAR数据集主要用于自监督预训练,帮助深度学习模型更好地理解冰川崩解前缘的动态变化,从而提高模型在崩解前缘提取任务上的性能。
SSL4SAR is an unlabeled SAR dataset consisting of 9,562 Sentinel-1 images and 14 Sentinel-2 images, which depict 14 Arctic glaciers with diverse sizes and calving front geometries. The dataset also includes one cloud-free optical image per glacier, captured by Sentinel-2 between June and August 2020. Developed to address the glacier calving front extraction task, SSL4SAR serves as a large-scale spatiotemporal database that provides calving front locations for subsequent analytical studies. Primarily designed for self-supervised pre-training, this dataset enables deep learning models to better comprehend the dynamic changes of glacier calving fronts, thus enhancing the models' performance on calving front extraction tasks.

- 1SSL4SAR: Self-Supervised Learning for Glacier Calving Front Extraction from SAR Imagery弗里德里希·亚历山大大学埃尔兰根-纽伦堡分校计算机科学系模式识别实验室 · 2025年



