search

shicaiwei123 / patch_based_cnn

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

Depth-based CNN missing from output folder #6

Open HKhawaja opened 2 years ago

HKhawaja commented 2 years ago

Awesome work on the implementation!! I saw that "depth_fasd.pth" is missing from the output/models folder. It is referenced in the test script: https://github.com/shicaiwei123/patch_based_cnn/blob/master/test/depth_cnn_test.py

HKhawaja commented 2 years ago

@shicaiwei123

shicaiwei123 commented 2 years ago

Hello, sorry for the late reply. The model depth_patch.pth has been uploaded just now. It's worth noting that, here, only the depth map estimation is implemented and the SVM has not been added. The current strategy used to classify is to calculate the mean value of the estimated depth map and compare it with the threshold (here is 0.5), if it is bigger than the threshold, it is classified as the true face, otherwise, it is the false face.

showbit01 commented 2 years ago

@shicaiwei123 is it only depth model or fusion of depth patch?

shicaiwei123 commented 2 years ago

The performance post on the github only leverage the patch.

But we also provide the code for depth model

At 2022-05-20 14:36:36, "Shobhit Sharma" @.***> wrote:

@shicaiwei123 is it only depth model or fusion of depth patch?

— Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you were mentioned.Message ID: @.***>

showbit01 commented 2 years ago

Yeah have implemented the depth with help of prnet module but again that is not surprising to see the overfitting results..

Ruhshan commented 1 year ago

Yeah have implemented the depth with help of prnet module but again that is not surprising to see the overfitting results..

Hi, can u help me in acquiring depth map with prnet? I tried this:

https://github.com/shicaiwei123/patch_based_cnn/issues/13

Not sure it's the correct way or not.

Thanks in advance.

showbit01 commented 1 year ago

Yeah have implemented the depth with help of prnet module but again that is not surprising to see the overfitting results..

Hi, can u help me in acquiring depth map with prnet? I tried this:

https://github.com/shicaiwei123/patch_based_cnn/issues/13

Not sure it's the correct way or not.

Thanks in advance.

https://github.com/clks-wzz/PRNet-Depth-Generation

use this