Python toolkit (Trinet-tools)¶
Trinet-tools is the open-source (MIT) Python toolkit for Trinet recordings. It can:
- parse
.imuand.vtsfiles into NumPy arrays; - repair recordings that some players or uploaders show as only a second long;
- extract motion data and timestamps from a video captured over USB;
- visualize a recording as video with live motion plots, and several kit cameras side by side on one timeline;
- export to MCAP for Foxglove and ROS 2;
- put card recordings on UTC time;
- package SD cards into upload-ready data deliveries;
- run stereo depth, motion HUDs and stereo-inertial odometry on stereo takes.
Install¶
git clone https://github.com/Panoculon-Labs/Trinet-tools.git
cd Trinet-tools
python3 -m venv .venv && source .venv/bin/activate
pip install -r requirements.txt
- Python 3 on Windows, macOS or Linux.
ffmpegandffprobemust be on yourPATHfor extraction and visualization (most package managers install both together).- The repair and SD-card ingest scripts need only the Python standard library.
Run the tests with pip install -r requirements-dev.txt && python3 -m pytest -q.
Quick start¶
from trinet_tools.reader import read_imu, read_vts, interpolate_imu_to_frames
imu = read_imu("recording1_1.imu")
vts = read_vts("recording1_1.vts")
print(f"{imu.num_samples} samples over {imu.duration_s:.1f} s at {imu.actual_rate_hz:.1f} Hz")
per_frame = interpolate_imu_to_frames(imu, vts) # motion data aligned to each video frame
print(per_frame[0]["frame_number"], per_frame[0]["accel"], per_frame[0]["gyro"])
Or start with the notebook
examples/tmf_metadata_and_calibration.ipynb:
it reads the metadata, motion data and calibration embedded in a real stereo recording shipped with
the repository and undistorts a frame.
Guides¶
- Inspect and repair — read, check and fix recordings
- Visualize and check sync — motion plots over video, kits side by side
- Export to MCAP / ROS 2 — one file for Foxglove and ROS 2
- Put recordings on UTC — using the app's wireless log
- Ingest SD cards — upload-ready deliveries with metadata
- Stereo and 3D — depth, motion HUD, VIO
Help and issues¶
Found a bug or a recording the toolkit can't read? Open an issue on GitHub.