Differences Between Running a Model in Caffe with C++ and Python

[Alfonso Soriano]

So basically, I have the same model running on the same program with one being implemented in C++ and one in Python. I am having an issue where the output is different between the two. I'm fairly certain the differences lie in the preprocessing of the image and how Python might do that implicitly but in C++ you have to do it explicitly, is anyone familiar with these differences and could point them out?

Thanks

[Alfonso Soriano]

Further, I have confirmed that the input in the data layer is exactly the same in both programs so that is not the issue.