BIG
收藏资源简介:
缺乏高分辨率图像分割数据集 是评估高分辨率图像分割模型的困难之一。为了解决这个问题,我们 呈现大数据集,一个具有50个验证和100个测试对象的高分辨率语义分割数据集。图像 大范围内的分辨率2048年 × 1600至5000 × 3600。 数据集中的每一幅图像都被仔细标记为 专业,同时保持与PASCAL相同的指导方针 没有空隙区域的VOC 2012。重新标记的PASCAL验证集和大数据集都可以在我们的 项目网站。
One of the primary challenges in evaluating high-resolution image segmentation models is the scarcity of high-resolution image segmentation datasets. To address this limitation, we present BigDataset, a high-resolution semantic segmentation dataset containing 50 validation and 100 test objects. The images have resolutions ranging from 2048 × 1600 to 5000 × 3600. Every image in the dataset has been meticulously annotated by professionals, adhering to the same annotation guidelines as PASCAL VOC 2012 with no void regions. Both the re-annotated PASCAL validation set and BigDataset are available on our project website.




