Sentinel-2 New Anomalies AI4QC

This dataset was used in the AI4QC project (Artificial Intelligence for Quality Control), in the context of the detection of new anomalies through unsupervised learning (unlabeled data). It consists of 6452 Sentinel-2 images (true color images in jpg format). The dataset was divided into training and testing folders (80% training and 20% testing). Two criterias were considered for the train/test split: seasonality and geographic location. An additional folder, "S2_additional_data" contains 61 more products which were flagged as anomalous by the MPC. This data is not included in the train/test folders but can be used if one wishes to increase the amount of S2 products.本数据集应用于AI4QC（Artificial Intelligence for Quality Control，质量控制人工智能）项目，适用于通过无监督学习（unsupervised learning）开展基于无标注数据（unlabeled data）的新型异常检测场景。该数据集包含6452张哨兵-2（Sentinel-2）影像，均为JPG格式的真彩色图像。该数据集被划分为训练集与测试集两个文件夹，其中训练集占比80%，测试集占比20%。本次训练测试集的划分遵循两项准则：季节性与地理位置。 新增文件夹"S2_additional_data"包含额外61份经MPC标记为异常的Sentinel-2影像产品。该数据未被纳入训练测试集文件夹，但如需扩充Sentinel-2影像样本量时可使用该数据。