search

eververas / 3DGazeNet

[ECCV 2024] 3DGazeNet: Generalizing Gaze Estimation with Weak-Supervision from Synthetic Views
https://arxiv.org/abs/2212.02997
57 stars 5 forks source link

Important files missing #3

Closed zhaishuyan closed 7 months ago

zhaishuyan commented 8 months ago

@Vagver, thank you for your work!

I have noticed that some crucial files are missing when running the code. The specific files are as follows:

xgaze/lms68_3D_all.pkl and xgaze/lms8_all.pkl mentioned in xgaze_preprocess.py. mpiiface/lms68_3D_all.pkl and mpiiface/lms8_all.pkl mentioned in mpiiface_preprocess.py. gaze360/lms68_3D_wbb.pkl, lms68_3D_wobb.pkl, lms8_all.pkl, paths_train_hp90.txt, and paths_test_hp90.txt mentioned in gaze360_preprocess.py.

Could you kindly provide the code for generating these files or provide detailed information about the data so that I can generate these files myself?

eververas commented 8 months ago

Hello @zhaishuyan,

Thank you for your interest in our work. We will provide download links for the files you mention soon.

For the time being you could employ the scripts that we already provide to generate the files. The “lms68_3D_all.pkl” files can be extracted using the “tools/preprocess_inference.py script” or your own 3D face landmarks localization algorithm. The “lms8_all.pkl” files can be generated by running the “inference.py” script on your images, using the provided checkpoint.