I have done a study about adding the features that you comment, checking the impact that could have adding the features to the model's training vector, and I have obtained the GoD if it could be relevant (as you obtain on your results), but adding the HSV vector does not seem to be, which is opposite with its graph of results, which in the graph (https://github.com/NarcissusHliangZhao/Face-Liveness-Detection/blob/master/results/6.svg) shows that LBP = 0.84, and LBP + HSV = 0.94 could you say how to pair these two resulting vectors (LBP and HSV for each side) and some more proof about the results obtained
The impact, I measure it in the vector differences of 100 images between real and fake, doing a tff, adding it and multiplying it with the conj.
parequena
commented
3 years ago
I have done a study about adding the features that you comment, checking the impact that could have adding the features to the model's training vector, and I have obtained the GoD if it could be relevant (as you obtain on your results), but adding the HSV vector does not seem to be, which is opposite with its graph of results, which in the graph (https://github.com/NarcissusHliangZhao/Face-Liveness-Detection/blob/master/results/6.svg) shows that LBP = 0.84, and LBP + HSV = 0.94 could you say how to pair these two resulting vectors (LBP and HSV for each side) and some more proof about the results obtained
The impact, I measure it in the vector differences of 100 images between real and fake, doing a tff, adding it and multiplying it with the conj.
Greetings, good toolbox.