OGLO Python SDK

Read touch and motion data from OGLO gloves. Record it, replay it, or capture it alongside webcam or OVISION video.

This is the 0.1.0rc8.dev2 development candidate. Managed firmware and JSONL recording are still under physical qualification; fleet rollout is not approved. See managed updates.

Requires Python 3.10+, firmware 0.9.10+ with schema 6, and a USB data cable.

Install

For automatic USB firmware updates on macOS/Linux, run this in the Python environment used by your collection program:

curl -fsSL https://github.com/OpenGraphLabs/oglo-python/releases/download/v0.1.0rc8.dev2/install.py | python - --auto-firmware

The same SDK serves every lab. Signed firmware is included; no glove list, separate firmware file or configuration editing is needed. Installation does not flash devices; compatible gloves are prepared at the next connect() or connect_pair(). See compatibility and disabling updates.

From this repository:

python3 -m venv .venv
source .venv/bin/activate
python -m pip install -e .

On Windows, activate with .venv\Scripts\Activate.ps1 in PowerShell. For a packaged build, use the download guide.

Quick start: record with OGLO Studio

Use Studio for a guided recording with two USB OGLO gloves (left and right) and one camera on the same computer. Plug them in with USB data cables and close any other app using the gloves or camera, including OGLO Viewer. The step-by-step Studio guide covers camera choices, calibration, review, exports, and troubleshooting. Use the localhost UI test checklist for a repeatable device-to-ZIP test.

From this repository, with the virtual environment activated (use Python 3.12+ and '.[studio,studio-ovision]' for native OVISION on Linux):

python -m pip install -e '.[studio]'
oglo doctor
oglo studio --output ./captures/studio

If oglo doctor reports a failure, fix the connection before recording.

Open http://127.0.0.1:8765/. If that port is in use, start Studio with oglo studio --port 8766 --output ./captures/studio and open http://127.0.0.1:8766/ instead. Studio is a Python app; no npm run dev is needed. Leave the terminal running while you record, then press Ctrl+C to stop it.

Follow the six steps in the page:

1. Connect and check. Select the camera and each glove, then click Connect selected devices. On macOS, choose OVISION left eye or OVISION right eye when available. Studio shows the selected 1920×1080 eye and saves the full packed 3840×1080 stereo video. Confirm the correct view appears, both glove frame counters advance, and a brief fingertip touch lights up each hand's taxels.
2. Calibrate. Choose Use saved calibration if it still matches the glove fit, or run a fresh five-second sweep. For a fresh sweep, repeatedly open and close both hands without touching anything, including your own fingertips. Studio shows a countdown, verifies both baselines, and displays the taxel spread and contact threshold. Check fingertip response before continuing.
3. Set up a button (optional). Test a keyboard-style USB pedal and map its keys, or continue with the on-screen controls.
4. Record. Enter a task description, click Start recording, perform the task, and click Stop. Wait for Studio to check the files.
5. Review. Select the take on the Review page, play its video, then choose Keep or Discard. Return to Record for another take if needed.
6. Export. Click Export kept takes, then Download dataset ZIP. The ZIP contains the video, camera timestamps, both gloves' sensor data, calibration, and checksums. The source episodes remain under ./captures/studio/, and the exported ZIP is also saved in ./captures/studio/exports/.

Mac OVISION capture keeps both eyes but only host timing, so it remains a source archive. For OG Center stereo+tactile post processing, run Studio on Linux, choose a native OVISION camera option, and select OG Center sensor source. That path saves native exposure timing, camera IMU, and calibration alongside both gloves, but physical Linux capture is still unverified. It requires task-content review and a separate OG Center import. See the Studio guide for pedal setup, data checks, and delivery limits. Scripts can use the same workflow through the collection SDK without starting the web page.

Read a glove

from itertools import islice
import oglo

with oglo.connect() as glove:
    print(glove.info.side, glove.info.channels)
    for frame in islice(glove.tactile(), 10):
        print(frame.counts)

Touch data has shape (5 fingers, 4 rows, 4 columns). Finger order comes from glove.info.channels. Values are sensor counts, not force in newtons.

Record and replay

import oglo

path = oglo.record("out/", seconds=60)
episode = oglo.replay(path)
print(episode.summary())

Recording writes JSONL directly in the backend's sensor format. Replay reads those same files. See recording for RAW/CLEAN handling.

Record with a camera

Connect camera and gloves to the same computer:

• USB webcam
• OVISION v1 — Linux, Python 3.12+

Send the whole output folder. The camera guides explain how to check it.

For repeated camera and glove episodes, use the Studio walkthrough above.

More

Calibration · Troubleshooting · All guides · Contributing · Security · License