Using Torch Framework to train and test colored and translated MNIST dataset
Pranav Aggarwal
Date - 4th December 2016
Note:
List of files
—Colored_train.lua
This program creates LeNet-5 model for colored background MNIST dataset
To execute
th Colored_train.lua
—Colored_test.lua
This program implements the model created in Colored_train.lua
To execute
th Colored_test.lua
—Translated_train.lua
This program creates LeNet-5 model for translated MNIST dataset
To execute
th Translated_train.lua
—Translated_test.lua
This program implements the model created in Translated_train.lua
To execute
th Translated_test.lua
To display images I have used interative torch which can be started by typing “qlua -lenv” on the terminal. Then used the following code to display the images
For example This will display the first 10 digits from the above mentioned dataset
require “image”
t7> d1 = torch.load(‘mnist-p1b-test.t7’)
t7> image.display{image=d1.data:narrow(1,1,10)}