Pupil Labs Camera

This repo contains functionality around the usage of camera intrinsics for undistorting data, projecting and unprojecting points.

It is mostly a wrapper around OpenCV's functionality, providing type hints, input validation, a more intuitive interface, and a few performance optimizations.

Installation

pip install pupil-labs-camera

or

pip install -e git+https://github.com/pupil-labs/pl-camera.git

Quick Start

The following code demonstrates how to use the library in the context of a Neon recording to undistort an image and 2D gaze points. Additional examples can be found in the example Jupyter Notebook.

import cv2

from pupil_labs import neon_recording as nr
from pupil_labs.camera import Camera

recording = nr.NeonRecording(
    "/path/to/recording"
)
camera = Camera(
    pixel_width=1600,
    pixel_height=1200,
    camera_matrix=recording.calibration.scene_camera_matrix,
    distortion_coefficients=recording.calibration.scene_distortion_coefficients,
)

data = zip(recording.scene, recording.gaze.sample(recording.scene.time))

for scene_frame, gaze_sample in data:
    distorted_image = scene_frame.bgr
    undistorted_image = camera.undistort_image(distorted_image)

    distorted_gaze = gaze_sample.point
    undistorted_gaze = camera.undistort_points(distorted_gaze)

    cv2.circle(
        distorted_image,
        tuple(map(int, distorted_gaze)),
        radius=25,
        color=(0, 0, 255),
        thickness=5,
    )

    cv2.circle(
        undistorted_image,
        tuple(map(int, undistorted_gaze)),
        radius=25,
        color=(0, 255, 0),
        thickness=5,
    )

    cv2.imshow("Distorted Image", distorted_image)
    cv2.imshow("Undistorted Image", undistorted_image)
    cv2.waitKey(30)