Open josecanedo opened 4 years ago
Strangely the last but one column has two ones, rather it should have a 1 and -1. I matched the matrix with standard MPII names I found it to be correct otherwise.
I noticed that in the discriminator these was a wassertein loss. But in the code it is defined as 'mean(y_true * y_pred)'. So there had any precess for it. @Aratrik @josecanedo
Strangely the last but one column has two ones, rather it should have a 1 and -1. I matched the matrix with standard MPII names I found it to be correct otherwise.
I count exactly one pair of 1 and -1 in each column. You may have counted along the row?
Excellent paper and code! Thanks.
By analyzing the KCS matrix and reassembling the skeleton following the paper i got something like: 6 -> 7 -> 8 -> 9 7-> 10 -> 11 -> 12 7-> 13 -> 14 -> 15 which seems to be the upper body starting at the root (hip joint), but the lower bones assemble like this: 0 -> 6 0 -> 1 -> 2 3 -> 4 -> 5 I was expecting something like this: 6 -> 0 -> 1 -> 2 6 -> 3 -> 4 -> 5
Is there a error in the KCS matrix or I misinterpreted the skeleton?
KCS matrix