clks-wzz

Which "gradient type" do you use here? https://github.com/clks-wzz/FAS-SGTD/blob/master/fas_sgtd_single_frame/util/util_network.py#L57 I recommend that you can try "gradient_type == 'type1'".

Maybe the dataset is not correctly written, so that the length of dataloader(test_input_fn) is zero. Please check it.

1. "Depthwise Spatial Gradient Magnitude" indeed plays an important role in our experiment because it can grasp the detailed spoofing clues. 2. What's the setting in your experiments, such as...

The mobilenet_v2 may be too deep to converge for RSGB and depth supervised learning. Which loss function do you try in your experiment, binary or depth? A pretrained model initialization...

We have tried deeper network architectures, and found that the deeper models even may be not suitable to the depth supervision task. You can try RSGB in the former layers,...

"video_frame_count" is zero. You should adjust the test file to your own style.

https://github.com/clks-wzz/FAS-SGTD/blob/master/fas_sgtd_single_frame/generate_data_test.py#L105 Maybe the pattern/style that you set OULU-NPU dataset is not adaptive to this python file. You can try change it for your own custom.

You should convert video format to frame format for evaluation. Generally speaking, the score of one video is the average of all/sampled frames' scores of the corresponding video.

> TNX, but I have another question, what is the structure of the test data folder? Do I need to put my video frames in a subfolder with a specific...

> > You should convert video format to frame format for evaluation. Generally speaking, the score of one video is the average of all/sampled frames' scores of the corresponding video....