mohammed-amr
commented
2 years ago Sorry, the dataset name is a isnomer. This class file is meant to be used with scenes captures using the ios-logger capturing app and processed with NeuralRecon's processing scripts.
Sorry!
Hi, Thanks for publishing this awesome work, great job! I am trying to use the Arkit scenes dataset with your amazing work. I found out that the Arkit scenes file format is different from the ScanNetv2 Dataset, especially for the camera pose.
The format for Arkit scenes(3dod): ARKitScenes/threedod/sample_data/40753679/ ├── 40753679_3dod_annotation.json ├── 40753679_3dod_mesh.ply └── 40753679_frames ├── color.traj # camera poses ├── color_intrinsics # camera intrinsics │ ├── 6845.7061.pincam # filenames are indexed by timestamps │ ├── 6845.8060.pincam │ ├── 6845.9060.pincam │ └── ... ├── depth_densified # depth frames │ ├── 6845.80601079.png # filenames are indexed by timestamps │ ├── 6845.90596450.png │ └── ... └── wide # color frames ├── 6845.70605696.png # filenames are indexed by timestamps ├── 6845.80601079.png ├── 6845.90596450.png └── ... color.traj format: Column 1: timestamp Columns 2-4: rotation (axis-angle representation in radians) Columns 5-7: translation (in meters)
Could you please tell me what is the correct input format for arkit_dataset.py?
Thank you!