Closed BeckZhang closed 4 years ago
Hey, Yes that is what I would expect as well. However, for some reason I was seeing plenty of precision difference when I used my MacOS based pretrained model on Windows. I had to retrain the model on Windows and it worked with similar precision after that. This was a surprising observation and couldn't dig into the actual root cause due to time limitation on my end.
Maybe you could compare the input data, for seeking the os difference. Just FYI.
I cant anymore. Dont have windows platform to try on now.
I think the model should be the same, regarding the training data set (in path imgfolder_b), in spite of the OS. So, what's the difference between models for mac and win?