Pretrained model (Nvidia DIGITS) making wrong classifications

[casiowatch]

Hi all, I am new to Caffe. Have trained a model using Nvidia DIGITS, with 6 distinct classes and ~100% classification accuracy as verified on DIGITS. Next, I tried running the model on this code: https://www.cc.gatech.edu/~zk15/deep_learning/classify_test.py, as well as the Jupyter Notebook from https://github.com/BVLC/caffe/blob/master/examples/00-classification.ipynb. However, when I tried to classify an image using them, they have consistently returned a predicted class of '5' with a fairly high confidence level (>90%). Seems like the classifier code is not correctly reading the labels from the trained model. Is there a way around this? Thanks in advance!

Cheers,

Casiowatch