hi, i'm really new to matlab and k-means clustering. i'm sorry for my question but i really get lost now. if anyone can help me or have matlab code how to use k-mans clustering using feature extraction as an input value so that i can get the centroids, please ? and also, i know the meaning of c1, c2, ..,cN is the clusters of the images. but how can i get those clusters using this k-means ?
i've already tried k-means in matlab but still i don't understand since the input is an image.
thank you so much
Hi Ade Aulya
MATLAB itself has a ready function called kmeans. Just give the image as a two-dimensional to input kmeans.
for example:
[cluster_idx, cluster_center] = kmeans(im_twodim,nColors,'distance','sqEuclidean', 'Replicates',3);
ofcourse in the following use kmeans for use k-means outputs (extracted features) as SVM inputs ...
https://datascience.stackexchange.com/questions/24324/how-to-use-k-means-outputs-extracted-features-as-svm-inputs
Hi Aulya,
The code depends on your task. You may need to give some info about the feature extraction technique that you managed to use.
Some feature extraction methods can be applied using cluster algorithms like "Expectation maximisation" or "EM" (i.e., density-based clusterer) to generate cluster membership values. Then, pass the result to K-means. However, other features extraction approaches, such as PCA, has to be applied first before clustering the data with K-means.
HTH.
Dr. Samer Sarsam
hello, sirs.
thank you all so much for your answers. i really appreciate it. but i still didn't get it. would you mind to give me an explaination, please sir ?
also, is this your statement ' features extraction approaches, such as PCA, has to be applied first before clustering the data ' means i should get the value (extract feature) of glcm from these images itself ?
if that so, i aready done with it, sir.
in this case, i used the GLCM values (such as contrast, correlation, energy, etc) as the input for k-means to get the cluster of images (like c1,c2,...,cN), what should i do, sir ? actually this clusters (c1,c2,..,cN) will i use to extract the cluster co-occurence matrix feature, that will be the input for machine learning such as SVM.
sorry for my lack of technical language. This is my first time programming in matlab and I'm quite terrible at. Not expecting a straight up answer fed to me, just really don't know how to proceed form here.
Aulya,
You can start working with GLCM first. Then, apply PCA (if necessary). Finally, run K-means clusterer.
Note that you could use, for instance, the Gabor filters that have several advantages over GLCM.
Dr. Samer Sarsam
already done with glcm, dr Samer Sarsam . that's why i continue now to k-means using those glcm value as the input of k-means.. also already done with k-means just still figuring out how this program can read multiple folders of those glcm output for once. do have any idea sir ?
thank you so much for your attention, sir.
best regards.
I'm not sure that I get your point, but K-means will group the instances that share similar features.
Dr. Samer Sarsam
- Badges
- Science topic