Hello! Thank you for your interest in 3DFacilities If you have not already, please go to the 3DFacilities website thomasczerniawski.com/3dfacilities and fill out the terms of use form. 3DFacilities is an annotated RGB-D dataset of building facilities. It currently contains over 25,000 individual frames and 110 scene reconstructions. Directory Structure 3DFacilities_Complete ROOT |_ 101 [scan] |_ 2D_color [2D color images] |_ 101_00001.png (Dimensions: 320 x 240, Bit depth: 24) | . | . | . |_ 101_00152.png |_ 2D_depth [2D depth images range 1-10m] |_ 101_00001.png (Dimensions: 320 x 240, Bit depth: 8) | . | . | . |_ 101_00152.png |_ 2D_labels_class [semantically segmented 2D image - class] |_ 101_00001.png (Dimensions: 320 x 240, Bit depth: 8) | . | . | . |_ 101_00152.png |_ 2D_labels_instance [semantically segmented 2D image - instance] |_ 101_00001.png (Dimensions: 320 x 240, Bit depth: 24) | . | . | . |_ 101_00152.png |_ 2D_meta [metadata for each 2D image] |_ 101_1.json |_ latitude |_ longitude |_ heading |_ acceleration |_ magnetometer |_ cameraGLProjection |_ gyroscope |_ cameraViewPoint |_ time | . | . | . |_ 101_152.json |_ 3D_labels_instance [semantically segmented 3D mesh - instance] |_ 101.ply (segmented 3D mesh) |_ 101.xml (mapping color to semantic label) |_ snapshot00.png |_ trajectory.log (trajectory of range camera during scan, i.e. camera pose) |_ 3D_mesh [textured 3D mesh] |_ Model.jpg |_ model.mtl |_ Model.obj (color 3D mesh) | . | . | (191 missing) | . | . |_ 210 Classes in the dataset are as follows. RGB colors listed correspond with colors in semantically segmented 2D images - class. unlabelled 0 0 0 furniture 0 51 0 door 0 51 102 wall 255 153 255 floor 102 0 102 window 0 0 102 ceiling 51 51 0 column 153 51 102 beam 102 0 204 stairs 255 153 0 railing 0 102 102 light fixture 255 255 204 elevator 51 102 255 plumbing 51 204 255 duct 0 255 0 diffuser 204 0 255 sprinkler 255 0 0 cable tray 255 255 0 conduit 255 153 102 background 128 128 128 In order to modify or create new 3D and 2D annotations, we refer you to https://github.com/scenenn/sese.