Skip to main content
Version: 1.x (legacy)

Find and read your output

Every recording, whether it's a calibration or a mocap session, lands in its own folder under ~/freemocap_data/recordings/. This page is a map of what's actually in there.

Where to look​

~/freemocap_data/recordings/
└── {recording_name}/
    ├── synchronized_videos/          # Frame-synchronized camera videos
    ├── annotated_videos/             # Videos with detection overlays
    ├── output_data/                  # 3D keypoints, center of mass, etc.
    │   ├── *.npy
    │   ├── *.csv
    │   └── *.parquet
    ├── logs/                         # Per-recording log files
    ├── {name}_camera_calibration.toml
    ├── {name}_recording_info.json
    └── {name}.blend                  # If you've exported to Blender

A migration to a slightly different layout (videos/synchronized/, videos/annotated/, and output/ instead of the names above) is in progress in the codebase; both layouts are recognized today, but the names above are what you'll actually see on disk right now. If a future version of FreeMoCap looks slightly different from this, that's why.

A few other defaults worth knowing:

WhatWhere
All FreeMoCap data~/freemocap_data/
All recordings~/freemocap_data/recordings/
Bundled test recording~/freemocap_data/recordings/freemocap_test_data/
Logs~/freemocap_data/logs_info_and_settings/logs/
Your most recent calibration~/freemocap_data/calibrations/last_successful_camera_calibration.toml

What's in each piece​

  • synchronized_videos/ is your raw footage, one file per camera, already time-aligned. annotated_videos/ is the same, with detected keypoints drawn on top, useful for checking tracking quality by eye.
  • output_data/ is where the actual mocap data lives: .npy arrays (see array shapes and units for exactly what's in each one), plus .csv and .parquet versions of the same data for tools that don't read .npy directly. See the output data model for what these numbers actually mean and how they were produced.
  • {name}_camera_calibration.toml is your calibration, in plain text. It's what reconstruction reads to know where each camera was.
  • {name}_recording_info.json records how the recording was made: camera settings at record time, and whether it was tagged as a calibration or a mocap recording. It's what makes a recording reprocessable later with different settings, without having to remember your original setup.
  • {name}.blend only shows up once you've exported to Blender; see Open your recording in Blender.

Next steps​