UG2
收藏资源简介:
UG2数据集是由圣母大学计算机科学与工程系创建的视频基准数据集,旨在评估图像恢复和增强技术对自动视觉识别的影响。该数据集包含超过150,000个标注帧,涵盖数百个ImageNet类别,用于评估图像伪影和其他条件对基于深度学习的对象分类方法的影响。UG2数据集包含三种不同的真实世界场景视频:无人机和有人驾驶滑翔机拍摄的非控制视频,以及地面控制视频。此数据集的应用领域包括机器人导航、场景重建、科学研究和视觉监控,旨在解决图像伪影对视觉识别算法的影响问题。
The UG2 dataset is a video benchmark dataset created by the Department of Computer Science and Engineering at the University of Notre Dame, aiming to evaluate the impact of image restoration and enhancement techniques on automated visual recognition. This dataset contains over 150,000 annotated frames spanning hundreds of ImageNet categories, and is used to assess the effects of image artifacts and other conditions on deep learning-based object classification methods. The UG2 dataset includes three distinct real-world scenario videos: uncontrolled footage captured by drones and manned gliders, as well as ground-controlled footage. The application scenarios of this dataset cover robot navigation, scene reconstruction, scientific research, and visual surveillance, and it is designed to address the issue of the impact of image artifacts on visual recognition algorithms.

- 1UG^2: a Video Benchmark for Assessing the Impact of Image Restoration and Enhancement on Automatic Visual Recognition圣母大学计算机科学与工程系 · 2018年



