Query: lm_reverse_list in model/dataset.py has 66 face landmark points specified
GayatriPurandharT opened this issue · comments
Hi, I have processed the data of SiW-M dataset.
The input folder contains live and spoof images- with cropped faces and 164 landmarks npy file.
I followed the same folder structure as provided in this repo.
problem:
IndexError: index 16 is out of bounds for axis 0 with size 1
Did you sample 66 points out of 148 or did you use some library which produces 66 points?
I used the face-alignment (https://github.com/1adrianb/face-alignment) that you suggested in a previous issue.
Please help me to choose the right landmark list to be able to train the model.
Thank you very much!
The error was due to the mismatch in dimension of my input npy face landmarks and the one expected in dataset.py. I had a 3d nympy array of landmarks. I converted back to 2d and that solved this issue.
Hi @GayatriPurandharT. I am also currently trying to replicate the results of this paper but I am unable to find the SiW-M dataset. Can you kindly share any pointers to the Siw-M dataset.