2 class classifier using caffenet

[rahmaniacc]

Hello

I modified the imagenet example to train a two class classifier and have a train.txt and val.txt that has close to 1000 images with  the labels 0 and 1 (about 60-40 split). I trained a net and get an accuracy of ~89%. But when i use a small python script to test the model, i get the following

Predicted class is #48001.

Is there some thing i am missing since 48001 is not a class lable i have defined. (Is there any file to define the classes other than the train.txt and val.txt files ?)

Also are there any suggestions for parameters to train a model for only 2 categories? I use a batch size of 256 and base learning rate of 0.001

Thanks