TrainFors
收藏资源简介:
TrainFors是一个大规模的图像篡改检测和定位训练数据集,由USC Information Sciences Institute创建。该数据集包含100万张图像,其中20万张为原始图像,80万张为经过四种篡改类型(图像拼接、复制移动伪造、移除伪造和图像增强伪造)处理的图像。数据集的创建旨在标准化图像篡改检测和定位任务的训练过程,通过使用真实世界的篡改实例来提高模型的性能。TrainFors的应用领域主要集中在图像取证,旨在解决图像篡改检测和定位的问题,以应对社会中日益增长的图像篡改和虚假信息传播问题。
TrainFors is a large-scale training dataset for image tampering detection and localization, developed by the USC Information Sciences Institute. This dataset comprises one million images, including 200,000 original images and 800,000 tampered images that have undergone four types of tampering operations: image splicing, copy-move forgery, object removal forgery, and image enhancement forgery. The dataset is designed to standardize the training workflow for image tampering detection and localization tasks, and improve model performance by leveraging real-world tampering instances. Its primary application domain is image forensics, aiming to address the challenges of image tampering detection and localization to cope with the escalating issues of image tampering and disinformation propagation in society.

- 1TrainFors: A Large Benchmark Training Dataset for Image Manipulation Detection and LocalizationUSC Information Sciences Institute · 2023年



