Skip to content

Latest commit

 

History

History
94 lines (73 loc) · 5.47 KB

CAPTURE.md

File metadata and controls

94 lines (73 loc) · 5.47 KB
Raw

Capture data format

The format is inspired by the Kapture format from Naver Labs but improves on the following points:

  • it can handle multiple sessions per location
  • it supports other assets: meshes, depth renderings, transformations between sessions, etc.
  • image matching is delegated to hloc
  • the API is optimized for each of use and a smooth learning curve

Overview

Here is an example of file structure for a capture at location1:

location1/                                  # a Capture directory
├── sessions/                               # a collection of Sessions 
│   ├── navvis1/                            # NavVis Session #1
│   │   ├── sensors.txt                     # list of all sensors with specs
│   │   ├── rigs.txt                        # rigid geometric relationship between sensors
│   │   ├── trajectories.txt                # pose for each (timestamp, sensor)
│   │   ├── images.txt                      # list of images with their paths
│   │   ├── pointclouds.txt                 # list of point clouds with their paths
│   │   ├── raw_data/                       # root path of images, point clouds, etc.
│   │   │   ├── images_undistorted/
│   │   │   └── pointcloud.ply
│   │   └── proc/                           # root path of processed assets
│   │       ├── meshes/                     # a collections of meshes
│   │       ├── depth_renderings.txt        # a list of rendered depth maps, one per image
│   │       ├── depth_renderings/           # root path for the depth maps
│   │       ├── alignment_global.txt        # global transforms between sessions
│   │       ├── alignment_trajectories.txt  # transform of each pose to a global reference
│   │       └── overlaps.h5                 # overlap matrix from this session to others
│   ├── hololens1/
│   │   ├── sensors.txt
│   │   ├── rigs.txt
│   │   ├── trajectories.txt
│   │   ├── images.txt
│   │   ├── depths.txt                      # list of depth maps with their paths
│   │   ├── bluetooth.txt                   # list of bluetooth measurements
│   │   ├── wifi.txt                        # list of wifi measurements
│   │   ├── raw_data/
│   │   │   ├── images/
│   │   │   └── depths/
│   │   └── proc/
│   │       └── alignment/
│   └── phone1/
│       └── ...
├── registration/                           # the data generated during alignment
│   ├── navvis2/
│   │   └── navvis1/                        # alignment of navvis2 w.r.t navvis1
│   │       └─ ...                          # intermediate data for matching/registration
│   └── hololens1/
│   │   └── navvis1/
│   └── phone1/
│       └── navvis2/
└── visualization/                          # root path of visualization dumps
    └─ ...                                  # all the data dumped during processing (TBD)

A Python interface that mirrors this file structure is available in scantools.capture. It can be created from an existing Capture file tree, but also create such tree from scratch and update it as it is modified.

File formats

Text files .txt contain utf-8 lines that are comma separated values (CSV) or comments starting with #.

Point clouds and meshes are in .ply PLY format, preferably in binary with float values.

Images are in PNG or JPEG formats.

Depth maps are 16-bit PNG images scaled by scantools.utils.io.DEPTH_SCALE.

Image features and matches are handled by hloc (currently HDF5 files).

Details

1. Session data

  • sensors.txt, rigs.txt, trajectories.txt follow the Kapture format. However, the pose convention is reverted: rigs.txt contains camera-to-rig transformations and trajectories.txt contains sensor-to-world transformations.
  • images.txt, pointclouds.txt, depths.txt, wifi.txt, and bt.txt follow the specifications of their corresponding records_*.txt in Kapture.

2. Processed files

  • proc/meshes/ contains all the meshes generated from the point cloud(s).
  • proc/depth_renderings.txt is a list of depth maps and their relative paths in proc/depth_renderings/
  • proc/alignment_global.txt is a list of global transforms w.r.t. other sessions with additional info (e.g. error statistics)
  • proc/alignment_trajectories.txt is a list of transforms w.r.t an absolute reference, following the format of trajectories.txt

3. Registration

The data generated for the pairwise and global alignment of all sessions is stored in capture/registration/. This includes images features and matches, registration confidences, errors statistics, and some visualizations. Each folder registration/session1/session2/ contains the alignment of session1 w.r.t session2.

4. Visualization

This directory contains images, meshes, and other assets useful to verify the ground truthing process.