DS5_DMC_synthetic_pipeline_data

The dataset contains 4.03 GB of images in png file format, taken at 31974 locations in a 3D model of a drinking water pipe network; the KWR test network. At each location 4 images are generated; a rendered camera view [idxxxx-image_render] of what the camera of the snake-like robot would see using its head camera. From that same viewpoint a normal map [idxxxx-image_normal] is generated that encodes pipe surface normal vectors in pixel color. And, from the same viewpoint, images of "ground truth" (pixels that belong to) ridges [idxxxx-image_ridge] and pipe branches [idxxxx-image_branch] are generated. These should be approximated by the ridge and branch detection models, based on the camera image in the robot. Furthermore, real camera images [idxxxx-image_render] are taken at 1161 locations in a pipe section test setup (clean pipes) and the accompanying image [idxxxx-image_branch] with "branch" classified pixels (manually identified) are provided. All images have a resolution of 512x512 pixels. A standard personal computer was used to generate the images from a 3D model.