GameIR
收藏资源简介:
GameIR数据集是由圣塔克拉拉大学和未来华为技术有限公司联合开发的大规模计算机合成基准数据集,旨在推动游戏内容图像恢复技术的研究。该数据集分为两个部分:GameIR-SR用于支持低分辨率到高分辨率的图像恢复,包含19,200对LR-HR图像;GameIR-NVS则用于多视角合成,包含57,600个高分辨率图像。数据集通过CARLA模拟器和Unreal Engine生成,不仅包括RGB图像,还提供了分割图和深度图等GBuffer信息,以辅助图像恢复。GameIR数据集的应用领域主要集中在云游戏解决方案中,通过提高图像质量和减少传输带宽,优化游戏体验。
The GameIR dataset is a large-scale computer-synthesized benchmark dataset co-developed by Santa Clara University and Huawei Future Technologies Co., Ltd., aiming to advance research on image restoration technologies for game content. The dataset is divided into two subsets: GameIR-SR, which supports low-resolution to high-resolution image restoration and contains 19,200 LR-HR image pairs; and GameIR-NVS, which is designed for multi-view synthesis and includes 57,600 high-resolution images. Generated via the CARLA simulator and Unreal Engine, the dataset not only provides RGB images but also offers GBuffer-related information such as segmentation maps and depth maps to assist image restoration research. The GameIR dataset is primarily applied in cloud gaming solutions, where it optimizes gaming experiences by enhancing image quality and reducing transmission bandwidth.




