Image Classification

[ulas alk]

Hi Everyone,

I want to group some graphs (image files) based on their patterns. I have a lot of images and want to see if I can group them based on their similarity. I saw that there are algorithms in Python and R. As a newbie in this area, I do not know if it is possible in Stata. I can create graphs again in Stata if it helps. I would appreciate any help and suggestions.

Best,
Ulas

[Phil Bromiley]

You'll increase your chances of a useful answer by following the FAQ on asking questions - provide Stata code in code delimiters, readable Stata output, and sample data using dataex.

Have you googled this? I found https://www.bu.edu/sph/files/2014/05/MorganCART.pdf but that is dated 2014 and this is an area that is progressing quickly. It is possible that there are user-written routines I do not know about. Entering findit classification in the command window brings up (among other things)

cart from http://fmwww.bc.edu/RePEc/bocode/c
'CART': module to perform Classification And Regression Tree analysis /
This program performs a CART analysis for failure time data. It / uses the
martingale residuals of a Cox model to calculate / (approximate) chisquare
values for all possible cutpoints on all / the CART covariates. / KW:

chaidforest from http://fmwww.bc.edu/RePEc/bocode/c
'CHAIDFOREST': module to conduct random forest ensemble classification
based on chi-square automated interaction detection (CHAID) as base
learner / Implements random forest ensemble classifier (Breiman, 2001; /
Machine Learning) using the CHAID (Chi-square automated / interaction


If these routines do not do what you want, you might contact the authors of these routines since they are likely to be better informed on this that most.

[Nick Cox]

One answer is simply that if you can represent "image files" by variables in a Stata dataset, then you're in business. It would seem simpler to work from the original data, but as Phil signals, you're not giving a clear picture of your goals. I doubt that Python or R is different in this respect. It's not images, but numerical representations of images, that are clustered, so far as I know about this.

[ulas alk]

Thank you very much for your answers. I could not understand the relation between suggested modules and my question. Maybe I need to be more clear. All I want to do is to cluster image files based on their similarity. The image files are 2D graphs and they have different patterns. I want to group them if the patterns are similar to each other.

[Leonardo Guizzetti]

Hi Ulas,

The first issue to deal with is abstracting data from those images. As far as I know, Stata is not designed for importing and analyzing images (at least not in any direct way). If you can use some method of extracting information from those images in a way that you can then make a traditional dataset, then you could use Stata for data analysis. This is the core of any image analysis method really, turning graphical information into abstracted data.