mnist-cluttered's Introduction

Cluttered MNIST Dataset

A setup script will download MNIST and produce mnist/*.t7 files:

luajit download_mnist.lua

Example usage:

local mnist_cluttered = require 'mnist_cluttered'
-- The observation will have size 1x100x100 with 8 distractors.
local dataConfig = {megapatch_w=100, num_dist=8}
local dataInfo = mnist_cluttered.createData(dataConfig)
local observation, target = unpack(dataInfo.nextExample())

mnist-cluttered's People

Contributors

Stargazers

Watchers

mnist-cluttered's Issues

Download source unavailable

http://yann.lecun.com/exdb/mnist is no longer available.
Is there another way to download mnist_cluttered.tar.gz?

Funny class vector when generating 3 digits?

Hi,

I was wondering if the following is the intended output?

I set opt.nDigits = 3 and targetFilling = "mark" , and generated an example using,

observation , class_vec = unpack(dataInfo_train.nextExample())

which was a 6 another 6 and a 9,

and got the class vector as,

th> class_vec
0
0
0
0
0
0
1
0
0
1
[torch.DoubleTensor of size 10]

I though this was a bit strange because two 6's were generated? I was just wondering if it should be,

th> class_vec
0
0
0
0
0
0
2
0
0
1
[torch.DoubleTensor of size 10]

In general, if I generate 3 digits, two of which are the same, the class vector always show one in each class? Is this intended? If so would it be possible to request a change in the code to give the class vector as the number of digits generated for each class?

A similar thing happens when I generate two digits, and both are the same?

Thanks a lot for your help,

Best regards,

Recommend Projects