Preparing image data, implementing locally connected layer and prediction in pylearn2

[vigneshwer dhinakaran]

Hello Pylearn users,

I have an image data set which I would like use to for training a DNN with locally connected layer and obtaining predictions for some test images.

As far I have understood I need to do the following,

1. make_datasety.py

2. train.py dataset.yaml

3. predict.py <model.yaml> <test_data.pkl> <predictions>

Below are my questions,

1. How to make a customized data set for my image dataset ?

2. In train.py for implementing locally connected layer how do I add it yaml file model ?

I failed to come across a complete example covering an end to end process.

I am very new to pylearn2, would be great if you guys could share some details in order for me to get started.