P2C-COD, T-COD
收藏资源简介:
P2C-COD和T-COD是由复旦大学计算机科学与技术学院开发的两个新型数据集,旨在支持弱监督伪装物体检测(WSCOD)任务。P2C-COD是一个点监督数据集,通过点标注提供精确的物体位置信息,而T-COD则是一个文本监督数据集,利用文本提示标签生成伪掩码。这两个数据集结合了点标注和文本提示的优势,能够生成高质量的伪标签,用于训练弱监督模型。数据集的应用领域包括野生动物保护、医学图像分割、战场敌情检测等,旨在解决伪装物体检测中的复杂视觉模式识别问题。
P2C-COD and T-COD are two novel datasets developed by the School of Computer Science and Technology, Fudan University, designed to support the weakly supervised camouflaged object detection (WSCOD) task. P2C-COD is a point-supervised dataset that provides precise object location information via point annotations, while T-COD is a text-supervised dataset that generates pseudo-masks using text prompt labels. These two datasets combine the advantages of point annotations and text prompts, enabling the generation of high-quality pseudo-labels for training weakly supervised models. The application scenarios of these datasets cover wildlife conservation, medical image segmentation, battlefield enemy situation detection and other fields, aiming to solve the complex visual pattern recognition problems in camouflaged object detection.




