# Clustering images based on timestamp

I want to make folders of images of users in a meaningful way. The images have only the timestamp of creation associated with them. Each folder can have a maximum of k images.

I can use median or mean as a way of deciding which images were taken in a definite time period.

What approach should I use to group these images in folders in such a way that the user is most intuitively able to make sense of the folders?

I am also asking here what would be an intuitive distinction for the user(I cannot use the day or hour of the image to group as I have the k images restriction.)

• Sorry but it seems that you have decision problem: which statistics are to be used for "clustering". There is no best option and using only timestamps gives: you can split by days, by hours (like 16-20 from everyday) or take $k$ elements from start or within hour... And there are more options. Have you considered at least using colors of the images?