pupil-labs / pye3d-detector

pye3d Detector

pye3d implements a published mathematical 3D eye model capturing ocular kinematics and optics (see Swirski and Dodgson, 2013, as well as Dierkes, Kassner, and Bulling, 2019.

Installation

We recommend installing the pre-packaged binary wheels from PyPI:

pip install pye3d

For more information see the documentation.

pye3d-detector's Issues

ModuleNotFoundError: No module named 'pye3d.cpp.pupil_detection_3d'

Dear author

I keep getting this error while running the code, but my environment is always configured. Is there any way to solve this problem

warning when debug window open.

I repeatedly get the following error when opening the debug window in pupil.

.../pye3d/detector_3d.py:603: RuntimeWarning: invalid value encountered in true_divide
  bin_data = bin_data / max_bin_level

It might be due to using python 3.8.5.

Some questions about focal length？

License question

Hello!
I'm looking for some clarification on the licensing and terms for this project - both by itself and as it relates to pupil. PyPI lists it as MIT + Proprietary (I only noticed the Proprietary part after I started developing with it). The MIT label looks like it got pulled in from a template, so that may have been unintentional? It doesn't appear in any of the components or files I've seen so far.

pupil, which has a dependency on pye3d

• "All source code written by Pupil Labs is ... [LGPL]"
• COPYING - GPL
• COPYING.LESSER - LGPL
• File headers - mostly LGPL, with one public domain and one BSD-3-Clause

pye3d itself

• LICENSE file - All Rights Reserved
• setup.cfg, pypi - MIT + Proprietary
• File headers - Most have no header, but a few like refraction_correction_cpp.h and background_helper.py have an LGPL header - I'm not sure how LGPL interacts with cpython, but that sounds like it could cause some problems if the library is intended to be proprietary?

How I can use my own camera model instead of Ideal camera model

In the code of readme
“camera = CameraModel(focal_length=561.5, resolution=[400, 400])”
But actually, the camera needs calibration.
so how set a new Camera parameters after calibration and undistort instead of Ideal camera model

Hello, I want to know the meaning of the output ellipse_3d in process_eye_video.py.

Hello, I test my own eye video use my own eye tracker, and I find the output of process_eye_video.py is the ellipse_3d, what is it mean in your paper? And how can I get the gaze estimation Vopt mentioned in your paper——A novel approach to single camera, glint-free 3D eye model fitting including corneal refraction？Thank you！

Using pye3d with non-human primates?

Hi, from this reply on discord, it seems I can change some constants from human standards to use with non-human primates. I'm not a python coder, so I am a bit lost on what I need to change and how to then use this in pupil core. I see this file https://github.com/pupil-labs/pye3d-detector/blob/master/pye3d/constants.py but don't understand what these constants refer to (human eyeball is 22 - 24mm diameter, so 11 - 12 mm radius, close to the 10.39 in that file, but what about the other constant?

Miguel on the forum suggested this code to edit: https://github.com/pupil-labs/pye3d-detector/blob/master/pye3d/detector_3d.py#L575 — it would be great if all the relevant constants could be defined separately and then included in the relevant places?

Access to Pye3d-Detector properties

Hi,
for now it's only possible to freeze the Pye3d-Model via the network API. It's the only property which is exposed, but I would also like to change the other parameter like karman_threshold oder swirski_threshold. Is it possible to implement this in future releases?
Best wishes
Matthias

Can I obtain or install the **plsd** library for image generation?

Can I obtain or install the plsd library for image generation? I want to generate some test pictures.

Something wrong when runing demo in ReadMe

detector_3d = Detector3D(camera=camera, long_term_mode=DetectorMode.blocking)
File "/Users/forrestzhu/anaconda3/lib/python3.7/site-packages/pye3d/detector_3d.py", line 136, in init
self.reset()
File "/Users/forrestzhu/anaconda3/lib/python3.7/site-packages/pye3d/detector_3d.py", line 187, in reset
ultra_long_term_model_cls=self._long_term_mode.value,
File "/Users/forrestzhu/anaconda3/lib/python3.7/site-packages/pye3d/detector_3d.py", line 213, in initialize_models
buffer_length=10,
File "/Users/forrestzhu/anaconda3/lib/python3.7/site-packages/pye3d/eye_model/base.py", line 47, in init
self.refractionizer = Refractionizer()
File "/Users/forrestzhu/anaconda3/lib/python3.7/site-packages/pye3d/refraction.py", line 19, in init
"radius", type, degree, custom_load_dir
File "/Users/forrestzhu/anaconda3/lib/python3.7/site-packages/pye3d/refraction.py", line 41, in load_config_from_msgpack
Refractionizer._validate_loaded_model_config(config_model)
File "/Users/forrestzhu/anaconda3/lib/python3.7/site-packages/pye3d/refraction.py", line 56, in _validate_loaded_model_config
raise ModelDeserializationError("Unrecognized format")
pye3d.refraction.ModelDeserializationError: Unrecognized format

Freeze eye sphere size

Hi! @papr
Is this the exact same 3d modes as the one used by Pupil Capture/Player?
I want to do pupillometry and look at how the pupil diameter reacts to light but then it is important that the eye-sphere size estimate does not change. In Pupil Capture/Player there is an option to freeze the model, can i freeze the model in pye3d as well?

Pye3d on eye videos of pupil invisible

Can this code be used to extract the 3d eye features for the eye videos extracted from pupil invisible hardware?

could you provide a demo/example or tutorials