Dense-Synthetic-View (DSynView)
收藏资源简介:
Dense-Synthetic-View (DSynView)是一个新的合成数据集,由德国凯撒斯劳滕-兰德大学和德国人工智能研究中心创建。该数据集设计用于评估球形密集文本到图像合成模型,包含1008个全景图像,这些图像通过组合三种背景文本提示和两种不同的前景提示集以及相应的自创掩码生成。数据集覆盖了室内房间、绿色田野和繁忙街道等场景,旨在解决球形图像生成中的布局控制和全景图像合成问题。
Dense-Synthetic-View (DSynView) is a novel synthetic dataset co-developed by the University of Kaiserslautern-Landau and the German Research Center for Artificial Intelligence (DFKI). It is designed to evaluate spherical dense text-to-image synthesis models, and consists of 1008 panoramic images generated by combining three background text prompts, two distinct sets of foreground prompts, along with their corresponding custom-made masks. The dataset covers diverse scenarios such as indoor rooms, green fields, and busy streets, and aims to tackle the challenges of layout control and panoramic image synthesis in spherical image generation.

- 1Spherical Dense Text-to-Image Synthesis德国凯撒斯劳滕-兰德大学,德国人工智能研究中心 · 2025年



