Python inference

Hi!

I have been trying to you your model in python in a inferece mode over another [dataset](http://pics.stir.ac.uk/2D_face_sets.htm) and I do not get the pain detection correctly.

I am using the trained weights that you provide in [google drive](https://drive.google.com/file/d/0B0OhXbSTAU1HUTVhNE1sX1o2STQ/view) and the graph that I use is [face_deploy.prototxt](https://github.com/happynear/PainRegression/blob/master/prototxt/face_deploy.prototxt) with a little change in the `InputData` layer by this:

```
layer {
name: "data"
type: "Input"
top: "data"
input_param {
shape: { dim: 1 dim: 3 dim: 112 dim: 96 }
}
transform_param {
mean_value: 128
scale: 0.0078125
}
}

```

I think that my python code is correct so I do not know what may be going wrong. Do you have any idea? Thank you