Classify objects in image

Hi,

I have a general idea however I dont know how to start.

I have large data set of images that each image includes many objects in it (some unusual items that I can find in google) that I dont know what are they.

I would like Matlab to categorize the objects from the images and then once they are categorized I will classify them and use the classified images to train the network for future images.

Any suggestions would be appriciated.

[Florian Morsch]

If you dont know what they are, how should MATLAB know? Depending on the objects you could try some pre-trained neural networks and if you lucky they give the correct output. But else this approach seems not very useful.

The image labeler app in MATLAB can be used to label images which then can be used for training, if that helps.

[ben]

I understood that there is an ootion to perform a classification even without knowing whats in the image by just finding equal features between the objects in the image where the features are pretty general as color,shape.

Then, I will go on ech group that it classified and determine what each group is.

[Florian Morsch]

You can seperate objects by color or shape, true. For example you could use https://de.mathworks.com/help/images/ref/imfindcircles.html to find circles (like coins) and extract them. If you know the general shapes and/or colors your objects have then you can use the simple image processing variants to extract them.

Those links might help you with that task:

https://de.mathworks.com/help/images/image-analysis.html This link provides several image processing examples with it.

https://de.mathworks.com/matlabcentral/fileexchange/15491-shape-recognition