about the toolbox's result of label and probability (is different from Caffe "classify.py")

[PieBoo]

Thank you for releasing this good toolbox.

I used this toolbox and found that the result label and probability of some images are much different from the result tested on "classify.py" (code in Caffe). Did someone meet the situation like this? What's the problem may causes this? And how does the label and probability calculated in Deep Visualization Toolbox. Thank you!