wang-lizhi / quantizationawaredeepoptics Goto Github PK

Is there a torch version of the code for this article?

In the paper the propagation step is represented by this formula :
$$U_{2}(x, y, \lambda)=F^{-1}{F{U_{1}e^{iz2\pi/\lambda}e^{-i*\pi*\lambdaz(f_{x}^{2}+f_{y}^{2})}$$
in the implementation, there is

        constant_exponent_part = np.float64(self.wave_lengths * np.pi * -1. * squared_sum)
        self.H = complex_exponent_tf(self.distance * constant_exponent_part,
                                     dtype=tf.complex64, name='fresnel_kernel')

I cannot find the code for $e^{iz2\pi/\lambda}$ part in the formula.
Could you please tell me where is it or if it can be ignored?