reproducing the VSRN paper

[DZZ95]

Hi, recently we are reproducing the VSRN paper.During the replication process, the running environment of the code was changed from Python 2.7 to Python 3.6, resulting in a much lower model accuracy.Do you have any good idea?Thank you very much!

[mesnico]

Hi @DZZ95, I noticed something similar. I solved this by using the same python version declared in their specifications (python 2.7), and the earliest possible pytorch version after the 0.4.1 (I think I used the 1.0.0, which was able to run on my system). With this configuration, I was able to obtain the claimed results on the test set using their pre-trained model.
Consider writing to the VSRN authors for more details on this issue.