Skip to first unread message
Michael Holm
Jun 28, 2016, 12:31:22 PM6/28/16
to Caffe Users
Dear Caffe Users Group,
I have two Googlenet trained caffe models that perform accurately on my linux platform where I trained them using DIGITS.
One of the models executes correctly on an Android platform, but the other model has all output scores between 49 and 51.
Can anyone suspect a cause for this problem?
Thank you,
Michael
I have two Googlenet trained caffe models that perform accurately on my linux platform where I trained them using DIGITS.
One of the models executes correctly on an Android platform, but the other model has all output scores between 49 and 51.
Can anyone suspect a cause for this problem?
Thank you,
Michael
Michael Holm
Jun 28, 2016, 12:48:30 PM6/28/16
to Caffe Users
The scores are on a scale of [0,100], and the behavior is repeated across a variety of Android versions. Caffe-android-lib is the fork of caffe used.
Reply all
Reply to author
Forward
0 new messages