search

MatteoZambra / Developmental-Approach-DBN

Support code
0 stars 0 forks source link

Links to datasets? #2

Open dvdplm opened 3 years ago

dvdplm commented 3 years ago

The Readme refers to "the MNIST dataset and the SZ dataset" without further specifications other than a link to a webpage with many download links.

Exactly what files are users expected to download?

MatteoZambra commented 3 years ago

MNIST dataset: automatically downloaded by torchvision data utils. You really only have to run the program with data_specs['stream'] set to 'create' and data_specs['save'] set to True, and you'll end up with the torch.Tensor dataset in the dataset folder.

SZ: Go to the page linked. The paragraph: "In order to try unsupervised deep learning on the prototypical cognitive modeling problem of visual numerosity perception investigated by Stoianov & Zorzi (2012), you can download the complete dataset of visual images here and follow the instructions provided inside the archive." Click on the here link, unzip and you'll find SZ_data.mat and SZ_data_test.mat. Save these two files in the dataset directory. Run the program with data_specs['stream'] set to 'create' and it will build the torch.Tensor dataset. No need to save it. In this case, the creation of the tensor dataset is not as expensive as the case of MNIST.

Hope it helps. Do not hesitate to write for further clarifications.

dvdplm commented 3 years ago

Click on the here link, unzip and you'll find SZ_data.mat and SZ_data_test.mat.

That link (http://ccnl.psy.unipd.it/research/visual-number-sense-dataset) downloads stoianovzorzi2012.tar which contains the following files:

➜  ~ ll Downloads/stoianovzorzi2012
total 10400
-rw-------@ 1 aggron  staff   1.4K Apr 12  2013 README.txt
-rw-------@ 1 aggron  staff   5.1M Apr 12  2013 StoianovZorzi2012_data.mat
-rw-------@ 1 aggron  staff   3.7K Apr 12  2013 stoianovzorzi2012_converter.m

Is running the stoianovzorzi2012_converter.m necessary or is the SZ_data_test.mat file sufficient?

As a side-note: the .m file extension is usually associated with Objective-C code so it might be helpful to users to add a note explaining that they need matlab/octave to run it.

dvdplm commented 3 years ago

Using GNU Octave, version 5.2.0, how do I run the converter script? When I run octave --traditional --verbose stoianovzorzi2012_converter.m I get a window showing a square with yellow squares on, but no .mat files are created in the folder. What am I doing wrong?

Note: without the --traditional flag I get a segfault.

MatteoZambra commented 3 years ago

No, it should be the .mat files already, let me search them. Indeed, the link I said does not point to data, I am sorry.

MatteoZambra commented 3 years ago

Updated. Now the link points to a OSF repo in which the supporting code and data for another work is saved.