negative age

[zhly0]

Hi，shamangary：
In the age prediction process，the age is negative，is there a way to avoid this？
Thanks。

[shamangary]

Hello @zhly0, in our funcion: https://github.com/shamangary/SSR-Net/blob/master/training_and_testing/SSRNET_model.py#L144
, the final prediction is 'age = (a+b+c)*V'

Considering the index is also controlled by the dynamic values 'index = (i+lambda_local*x[6][:,i])',
it is possible to cause negative values which is not supposed to happen. The solution could be:
(1) set lambda_local = 0 (2) change the initial index 'https://github.com/shamangary/SSR-Net/blob/master/training_and_testing/SSRNET_model.py#L301': from 0 to 1 and do the same in the three for loop.

Note that you need to re-train the model to make these changes work. Cheers.

[zhly0]

Thanks，I will try！