pupilio is a lightweight Python SDK for the Pupil.IO eye-trackers developed by Hangzhou Shenning Technology Co., Ltd. It offers a user-friendly interface that covers core eye-tracking functions like data recording, calibration, validation, and real-time access to gaze data. "pupilio" seamlessly integrates with popular platforms such as PsychoPy and PyGame, it enables quick development of eye-tracking studies.

The Pupil.IO eye-tracker

Pupil.IO is a high-speed, high-precision eye-tracking system featuring an all-in-one (AIO) plug-and-play design that is ideal for both scientific research and clinical applications. With minimal setup (just power on and start tracking), it delivers lab-grade accuracy in a compact, user-friendly form factor.

• Precision Tracking: Capture high-frequency eye movement and pupil dynamics with lab-grade accuracy.
• Seamless Compatibility: Native integration with PsychoPy, PyGame, and other Python experimental platforms.
• Multi-Modal Synchronization: Native support for LabStreamingLayer (LSL) to synchronize gaze, pupil dynamics, and event markers with EEG, fNIRS, and EMG via LabRecorder.
• Intuitive Workflow: Simplified calibration, validation, and recording with minimal setup.

Specifications

¹ DVS-2K edition available in Q4 2026. Final specifications subject to change.
² S2S-RMS, measured using artificial eyes with the filter turned on/off.
³ Measured against an industry-standard eye-tracker operating at 2000 Hz.
⁴ When operating at 200 Hz and both eyes are reliably tracked at a camera-to-eye distance of 70 cm.

Installation

Currently, all eye trackers shipped with Pupilio are pre-configured with the necessary Pupilio Python packages. If we need to upgrade this package, please enter the following command in the command prompt or PowerShell window:

pip install -U pupilio

Quick Start

Here is a simple example to get started with Pupilio:

# load PyGame library
import pygame
from pygame.locals import FULLSCREEN, HWSURFACE

# initialize PyGame
pygame.init()
# define the scree size
scn_width, scn_height = (1920, 1080)
# set a fullscreen mode with the above size
win = pygame.display.set_mode((scn_width, scn_height), FULLSCREEN|HWSURFACE)

# load Pupilio library
from pupilio import Pupilio

# instantiate a connection to the tracker
pupil_io = Pupilio()

# create a task session, and set a session name. 
# The session name must contain only letters, digits or underscores without any special characters.
pupil_io.create_session(session_name="quick_start")

# calibration and validation (recommended)
# set 'validate' to True if we would like to verify the calibration results
pupil_io.calibration_draw(screen=win, validate=True)

# start retrieving gaze
pupil_io.start_sampling()

# hang the main thread for 5 seconds by game
# eye tracking sampling are running on the background thread
pygame.time.wait(5 * 1000)

# stop eye tracking sampling
pupil_io.stop_sampling()

# sleep for 100 ms to capture ending samples
pygame.time.wait(100)

# save eye movement data
pupil_io.save_data("eye_movement.csv")

# release the tracker instance
# clean up Pupilio resources
pupil_io.release()

# quit pygame
pygame.quit()

Multi-Modal Synchronization (LabStreamingLayer)

Pupilio natively supports LabStreamingLayer (LSL) to synchronize eye-tracking data (gaze, pupil diameter, triggers) with EEG, fNIRS, and other biosensors via LabRecorder.

from pupilio import Pupilio, DefaultConfig

config = DefaultConfig()
config.enable_lsl = True
config.lsl_stream_mode = "standard"  # 12 channels (or "full" for 39 channels)

pupil_io = Pupilio(config=config)
pupil_io.create_session("eeg_eyetracking_session")

# Start sampling (LSL starts broadcasting automatically)
pupil_io.start_sampling()

# Forward numeric triggers to both LSL continuous Gaze (channel 11) & discrete Markers stream
pupil_io.set_trigger(101)

# Broadcast custom semantic string annotations
pupil_io.send_lsl_marker("STIMULUS_ONSET")

# Stop sampling (LSL stops automatically)
pupil_io.stop_sampling()
pupil_io.release()

For detailed stream specifications and integration with EEGLAB / MNE, see the LabStreamingLayer Integration Guide or GitHub Wiki.

Preprocessing for eye tracking data

Detect fixations and saccades using the I-DT (Identification by Dispersion Threshold) algorithm. The I-DT algorithm identifies fixations as groups of consecutive samples where the dispersion (sum of x and y ranges) remains below a specified threshold for a minimum duration.

Example:

import glob
import os

from pupilio import EventDetection

ed = EventDetection()

out_dir = 'output'
os.makedirs(out_dir, exist_ok=True)
f = open('error.txt', 'w')
for i in glob.glob(r'example.csv'):
    try:
        ed.detect(i, output_dir=out_dir, which_eye='right')
    except Exception as e:
        f.write(i + ',' + str(e) + '\n')
        f.flush()

Documentation

Comprehensive documentation is available at Pupilio Documentation.

Support

If you encounter any issues or have questions, please open an issue on GitHub or contact zhugc2016@gmail.com.

License

Pupilio is a proprietary software developed by Hangzhou Shenning Technology Co., Ltd. All rights reserved. Unauthorized use, distribution, or modification is prohibited without explicit permission. For licensing inquiries, please contact zhugc2016@gmail.com.

Acknowledgments

Pupilio is developed and maintained by Hangzhou Shenning Technology Co., Ltd. Special thanks to the community for their valuable feedback and support.