LivingOptics/hyperspectral-orchard

该数据集名为Living Optics Orchard Dataset，包含435张在英国最大的果园之一使用Living Optics相机拍摄的图像。数据集包括RGB图像、稀疏光谱样本和实例分割掩码。数据集来源于44个原始文件，对应435帧图像，因此多帧图像可能来自同一个原始文件。为了避免数据泄露，数据集采用8:2的比例在原始文件级别进行划分，确保训练集和测试集之间没有重叠信息。数据集包含3,785个皇家嘎啦苹果实例、2,523个梨实例和73个考克斯苹果实例，总计6,381个标注实例。未标注的光谱数据可用于分类器的负采样。数据集还提供了额外的未标注数据，可根据请求获取。

The Living Optics Orchard Dataset contains 435 images captured in one of the UKs largest orchards using the Living Optics Camera. The data includes RGB images, sparse spectral samples, and instance segmentation masks. The dataset is derived from 44 unique raw files corresponding to 435 frames, with an 8:2 split strategy performed at the raw file level to avoid data leakage. It contains 6,381 labelled instances, including 3,785 Royal Gala Apples, 2,523 Pears, and 73 Cox Apples. Unlabelled spectra can be used for negative sampling when training classifiers. The training dataset contains 3 classes: Cox apple, Royal Gala apple, and pear, with the remaining 1,855,755 spectra unlabelled and considered as a single background class.