DeepArt
收藏资源简介:
DeepArt数据集由重庆邮电大学自动化学院的研究团队创建,包含301幅由现代中国艺术家Feng Zikai手绘的诗画作品及其对应的GPT-4生成的图像。该数据集通过精心挑选的诗画作品,展示了GPT-4在图像合成中的纹理特征保真度。创建过程中采用了基于GPT-4的‘编码-解码’映射方法,以生成与原始数据相对应的合成数据。DeepArt数据集主要用于推动AI生成内容(AIGC)的保真度研究,特别是在多模态大型模型如GPT-4的图像生成能力评估中。
The DeepArt dataset was created by a research team from the School of Automation, Chongqing University of Posts and Telecommunications. It contains 301 hand-painted poetic painting works by Feng Zikai, a modern Chinese artist, along with their corresponding images generated by GPT-4. This dataset showcases the texture feature fidelity of GPT-4 in image synthesis through carefully selected poetic painting works. During its creation, a GPT-4-based "encode-decode" mapping method was adopted to generate synthetic data corresponding to the original dataset. The DeepArt dataset is primarily used to advance the research on the fidelity of AI-generated content (AIGC), especially in the evaluation of the image generation capabilities of multimodal large models such as GPT-4.

- 1DeepArt: A Benchmark to Advance Fidelity Research in AI-Generated Content重庆邮电大学自动化学院 · 2023年



