Street-Level Videos
The monocular data was gathered around TUM with a GoPro Hero 11 camera mounted in a car and handheld. The data was used for 3D reconstruction deploying neural radiance field for street view scenarios using a single monocular camera. Instructions, implementation, and more at the dedicated repository and in the thesis
Download the data here
From the left: Raw image, refined image, rendered depth, and rendered normals see the method and more visuals
Street-Level Facade Images
To support the validation of automatic texturing processes, perspective terrestrial optical images were manually captured using a Sony α7 digital camera. The image acquisition campaign was strategically planned to cover building facades with the minimum number of images per triangle, thereby ensuring consistent texture quality while avoiding the need for additional image stitching.
In addition, panoramic images from Google Street View were collected to test automatic image projection techniques on the building models.
Download the data here