search

hysts / pytorch_mpiigaze

An unofficial PyTorch implementation of MPIIGaze and MPIIFaceGaze
MIT License
346 stars 85 forks source link

how to make dataset #63

Closed ChiwoongLEE closed 2 years ago

ChiwoongLEE commented 2 years ago

Hello, thank you for sharing your valuable dataset.

I chose mpiigaze dataset as my first deep learning model as an undergraduate.

As a result of analyzing the mpiigaze data set, it was confirmed that various data and png files consisting of mat files were included.

I want to make the same as mpiigaze dataset with eye tracker device called pupil core. Can you give me some simple advice on how to make a dataset?

Also , as input data, can I build a dataset with only 3d gaze point and head_pose extracted with eye tracker equipment?

hysts commented 2 years ago

Hi, @ChiwoongLEE Sorry for the confusion. I'm not the author of the MPIIGaze paper and this is an unofficial implementation. About your questions, I think you should contact the authors. You can find contact information in their project page.

ChiwoongLEE commented 2 years ago

oh sorry thank you very much

Is the only dataset you used to train mpiigaze in Data/nomalized?

I am curious about the role of the annotation subset and evaluation subset folders in the MPIIGAZE dataset. I don't understand what is described on the official site.

hysts commented 2 years ago

@ChiwoongLEE

Is the only dataset you used to train mpiigaze in Data/nomalized?

Yes.

I am curious about the role of the annotation subset and evaluation subset folders in the MPIIGAZE dataset. I don't understand what is described on the official site.

It seems they are described in section 3.3 of their paper. For further detail, I recommend that you ask the authors directly.

ChiwoongLEE commented 2 years ago

thank you for your advise