DIML/CVL RGB-D Dataset
收藏资源简介:
DIML/CVL RGB-D Dataset包含200万张室内外自然场景的RGB-D图像。该数据集由室内数据和室外数据两部分组成,室内数据使用Microsoft Kinect v2采集,室外数据则使用ZED立体相机和内置立体相机采集。数据集规模庞大,包含100万室内数据和100万室外数据,特别之处在于室外数据采用手持立体相机采集,并提供置信度图以量化深度图的准确性。创建该数据集旨在推动基于深度学习的深度相关应用的发展,解决室内外场景深度信息获取的问题,适用于深度估计、RGB-D特征学习、室外场景理解和去雾等多种计算机视觉和图像处理任务。
DIML/CVL RGB-D Dataset contains 2 million RGB-D images of natural indoor and outdoor scenes. This dataset consists of two parts: indoor data and outdoor data. The indoor data was collected using a Microsoft Kinect v2, while the outdoor data was captured with a ZED stereo camera and built-in stereo cameras. With a large scale, it includes 1 million indoor samples and 1 million outdoor samples. A notable feature of this dataset is that the outdoor data was collected via handheld stereo cameras, and confidence maps are provided to quantify the accuracy of the depth maps. This dataset was developed to advance deep learning-based depth-related applications, address the challenge of acquiring depth information in indoor and outdoor scenarios, and is applicable to various computer vision and image processing tasks such as depth estimation, RGB-D feature learning, outdoor scene understanding, and dehazing.




