inputs_clone used error

[eeric]

in bifpn.py, the 196th line, as following:

1. w2[2, i] * inputs_clone[i + 1])/(w2[0, i] + w2[1, i] + w2[2, i] + self.eps) inputs_clone should be replace for pathtd. the 189th line, as following:
2. pathtd[i - 1] = (w1[0, i-1]pathtd[i - 1] + w1[1, i-1]F.interpolate( pathtd[i], scale_factor=2, mode='nearest'))/(w1[0, i-1] + w1[1, i-1] + self.eps) pathtd should be replace for inputs_clone.

[tadbeer]

Hi @eeric I am trying to train this on my own dataset, but cant achieve a successfull training, that is, the MAP is not increasing with increasing epochs. Have you tried training on your own dataset? Could you kindly get in touch with me?

[eeric]

@tadbeer，I modified this .py file, and antispoof model was trained, its effect of antispoof is better when model size is smaller,0.68M parametes, while model size is bigger, its effect is worse.