ScanNeRF

ScanNeRF数据集是由艾克安.ai和博洛尼亚大学合作创建，旨在为神经辐射场（NeRF）和神经渲染（NR）框架提供首个真实基准。该数据集通过一个成本低于500美元的自制扫描站，在5分钟内收集约4000张真实物体的图像，具有多个训练/验证/测试分割，用于评估不同条件下的NeRF方法性能。数据集内容包括高质量图像、像素掩码对象和控制重复的相机姿态，适用于研究神经渲染领域。ScanNeRF的应用领域包括虚拟现实和增强现实，旨在解决从真实世界到虚拟世界的无缝对象传输问题。

ScanNeRF Dataset was co-created by Ikang.ai and the University of Bologna, aiming to provide the first real-world benchmark for Neural Radiance Fields (NeRF) and Neural Rendering (NR) frameworks. This dataset captures approximately 4,000 real-world object images within 5 minutes via a self-made scanning station costing less than $500, and offers multiple train/validation/test splits for evaluating the performance of NeRF-based methods under various conditions. The dataset includes high-quality images, pixel-wise object masks, and camera poses with controlled repetitions, which are suitable for research in the field of neural rendering. The application domains of ScanNeRF cover virtual reality (VR) and augmented reality (AR), and it aims to address the problem of seamless transfer of real-world objects into the virtual world.