[GitHub] lichen11 commented on issue #14021: Load mxnet model trained in R to Python errors imply inconsistency in R and Python or bug in mx.io.ImageRecordIter

[GitBox <gi...@apache.org>]

lichen11 commented on issue #14021: Load mxnet model trained in R to Python errors imply inconsistency in R and Python or bug in mx.io.ImageRecordIter
URL: https://github.com/apache/incubator-mxnet/issues/14021#issuecomment-458711890
 
 
   I figured out what went wrong. The problem was that function mx.io.ImageRecordIter taking scale = 1/255 will produce all 0s on the images. I fixed the problem by setting scale = 1/255. and am able to achieve approx the same classification results. 
   I think this issue I raised also confirmed that 1. mxnet models trained in R can be loaded in Python 2. although mx.io.ImageRecorditer is channel first in Python and channel last in R, it does not affect the prediction results. Hence it is not necessary to add channel last in Python and vice versa. 

----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on GitHub and use the
URL above to go to the specific comment.
 
For queries about this service, please contact Infrastructure at:
users@infra.apache.org


With regards,
Apache Git Services