garrying/VMD-D
收藏资源简介:
VMD-D(视频镜面检测数据集)是第一个用于视频镜面检测的大规模数据集,包含14,987张图像帧,来自269个视频,每帧都有手动标注的二进制镜面掩码。视频被分割成剪辑,每个剪辑是一个独立的序列段。数据集分为训练集和测试集,训练集包含144个剪辑和7,836帧,测试集包含127个剪辑和7,151帧。每个样本包含四个列:image_id、clip_id、image和mask。数据集的原始磁盘布局和加载方式也在README中详细描述。
VMD-D (Video Mirror Detection Dataset) is the first large-scale dataset for Video Mirror Detection, containing 14,987 image frames from 269 videos with corresponding manually annotated binary mirror masks. Videos are split into clips, and each clip is an independent sequence segment. The dataset is divided into training and test sets, with the training set containing 144 clips and 7,836 frames, and the test set containing 127 clips and 7,151 frames. Each sample has four columns: image_id, clip_id, image, and mask. The original on-disk layout and loading methods are also described in detail in the README.




