search

shicaiwei123 / patch_based_cnn

the implement of Face Anti-Spoofing Using Patch and Depth-Based CNNs
30 stars 12 forks source link

Need guidance for acquiring depth map with PRNET #13

Open Ruhshan opened 1 year ago

Ruhshan commented 1 year ago

Hi, First of all, this repo is a gem. Truly helped me to better understand the paper. Already tried the end to end flow with patch based cnn using casia dataset.

But can u please help me which options/parameters you have used to acquire depth map with PRNET?

After some struggle I managed to run the following command:

python demo.py -i TestImages -o out --isShow True --isDlib True

Producing results like this:

image

And if I enable --isDepth flag I can get an output like this:

image

I have provided spoofed face but the 3d map doesn't look like one. Any direction?

Thanks in advance.

avinashsinghgit commented 4 months ago

Hey Ruhshan,

It’s great to see that someone as observant as you has pointed out the problem occurring in the last phase of the facial recognition project I’m currently studying.

I would be highly thankful if you could please share your solution or suggest an alternative approach to address this issue. Your input would be incredibly valuable to my learning.

Regards, Avinash Singh