Cannot replicate the same Circular Correlation Coefficients results showed in the paper when I train the model

[JimenezBarreroDavid]

When I train the model, I obtain similar results in the Pearson Correlation Coefficients for Distance and Phi, however for the Circular Correlation Coefficients I obtain very different results. I used this for calculating the coefficient after recovering the classes' value and converting it to radians. Am I missing something for calculating the coefficient? Or do you use some trick for training the angles? Perhaps can you share the code for calculating the Circular Correlation Coefficients?