can we have the pretrained checkpoint?

pydsgz / DeepVOG

Pupil segmentation and gaze estimation using fully convolutional neural networks

GNU General Public License v3.0

153 stars 68 forks source link

can we have the pretrained checkpoint? #12

Open shamanez opened 4 years ago

mangotee commented 4 years ago

Hi there, thx for reaching out. I am not sure what you mean with "pretrained checkpoint". In case you mean the pretrained network and weights, they are already available in this repo under: deepvog/model/DeepVOG_weights.h5. They can be loaded with the convenience function load_DeepVOG() under /deepvog/model/DeepVOG_model.py

shamanez commented 4 years ago

yeah, I got it. Sorry for spamming :)

Any idea, what if we test this with similar eye images taken from an RGB camera?

Is your camera is IR?

mangotee commented 3 years ago

Sorry for the late reply! Color images should work as well (internally, they get converted to grayscale for processing), and the performance should not be too bad, see this (3rdparty) blogpost: https://blog.orikami.nl/more-than-meets-the-eye-pupil-c53d4d3b1be3