GreatSplicing
收藏资源简介:
GreatSplicing是由重庆邮电大学创建的一个高质量手动合成数据集,包含5000张拼接图像,覆盖335个不同的语义类别。该数据集通过Adobe Photoshop手动创建,确保了拼接痕迹的真实性和高分辨率。GreatSplicing旨在帮助神经网络更好地学习拼接痕迹,提高跨数据集检测能力,适用于图像认证领域,解决现有数据集在语义多样性不足和检测模型过拟合问题。
GreatSplicing is a high-quality manually synthesized dataset developed by Chongqing University of Posts and Telecommunications. It contains 5,000 spliced images covering 335 distinct semantic categories. Manually created via Adobe Photoshop, the dataset ensures the authenticity of splicing traces and high-resolution quality. The dataset aims to help neural networks better learn splicing traces and improve cross-dataset detection capabilities, which is applicable to the field of image authentication and solves the problems of insufficient semantic diversity and detection model overfitting in existing datasets.

- 1GreatSplicing: A Semantically Rich Splicing Dataset重庆邮电大学 · 2023年



