I have images which have saved in the desk. The data saved as follow: 4 main folders (1,2,3 and 4) each folder has 26 subfolders ( these subfolders represent the class of images (A, B, C, D, ..,Z)). Each of these subfolder contains more than 500 images. However, I am looking for file or code in torch that can read these images. In MATLAB I could wrote a code but here I find it confuse. Could you please advise me.
What you can do is use Penlight (the library is installed when you install Torch).
Penlight provides pl.dir
that makes it easy to scan files in (sub-)folders. For example what you can do is:
local pl = require('pl.import_into')()
local t = {}
for i,f in ipairs(pl.dir.getallfiles('/data/foo', '*.jpg')) do
t[i] = { f, pl.path.basename(pl.path.dirname(f)) }
end
This creates a list of pairs (filename, class label = "A" or "B" ...). Of course you are free to change the file pattern (*.jpg
) or to omit it (in such a case Penlight will simply list all files). You can also load the images on the fly:
t[i] = { image.load(f), pl.path.basename(pl.path.dirname(f)) }
Or do that right after when manipulating t
.