Find Visual Duplicates For 1,700 Images

[Darius1968]

I've just acquired 1,700 new jpeg images from a friend.  I know that my database does not have any binary duplicates of these, but I'm almost sure that some are visual duplicates.  With this in mind, what is the best strategy for searching out my database for the visual duplicates?  I want the best settings that will yield matches for visual duplicates only, but can possibly be different dimensions, but nothing that is just similar. 

[Mario]

Select the 1700 files in a file window.
Search menu > Visually Similar Images.
Enable "Favor images with same orientation"
OK.

[Darius1968]

What about the check boxes to find with respect to similar colors or to search by the shape and color layout, and the slider for shape importance.  Where should these be to narrow the scope to visually duplicate and not similarity?  Also, I seem to remember IMatch 3.6 sorting the result of visually duplicates with those that were most similar at the top, and least similar, at the bottom.  The % of similarity was stated as well.  Is this possible in IMatch 5?  With this said, I see that it's possible to sort by similarity, but I also was wondering how to see the actual similarity as a percentage. 

[Mario]

Did you try out what I explained above? This should answer your questions. Should take only a few seconds.

Please see the IMatch help on the Visual Query features and the Result Window for answers to your other questions.

[Wolfgang]

I have the same kind of questions as Darius1968. When I look for similar images, I get a large quantity of files (50 as minimum, I saw no possibility to decrease this number), and - in my case - only one would be a very close fit with a high similarity. I was sucessful to sort by similarity and to show the similiarity value in the result window layout. But I could not figure out, how to decrease the number of results. In IMatch 3.6 we had the possibility to filter the result window and to "Only show images with at least this similarity" and we could choose for example 99 %. In the IMatch 3.6 help we could find a considerable amount of help items related to the key word "similarity", but in IMatch 5 I'm a bit lost in this respect. In IMatch 3.6 we had a sample script "Bookmark all files with a similarity of 99+ % in the current result set" written by Mario, which I used often. Do we have any similar script available in IMatch 5 ?

Often I modify pictures and/or decrease the size for example for printing or display on an iPad. I would like to locate all this pictures, which for the most part are very similar. Since all the files are mostly JPGs, I cannot use the Master/version approach, Mario strongly recommends to use it with different file types only.

Thanks for any help !

Wolfgang

[Mario]

But I could not figure out, how to decrease the number of results.

To keep things simple, the minimum number of matches returned for each original is 50. When you use the Default sort profile, the best matches come first.
Each Original builds a group with it's matches. You can expand collapse these groups as usual (see the File Window help) to see only one original and it's matches.

This easily allows you to select all matches (Ctrl+A), or do whatever you want to do with the similar files.

[Wolfgang]

Thanks for the answer. I'm not sure if I understand correctly what you mean by expand collapse these groups. I looked in the help, but could not find the reference how to expand collapse these groups. I see the small triangles on the side of each original and each group of 50 matches. This allows to reduce the amount of images shown, but only with a click on each of these triangles for each group by hand. I found under the help for the Result window the possibilty to select only the images for a single group by Ctrl + Alt + A. But then I still have 50 images to handle. I want to select/filter only the few images with the highest similarity, let's say above 99 % (as I could in IMatch 3.6).

I could not locate the right filter panel for filtering the results window on the Similarity value. I tried to apply the attribute filter to the results, but could not select the Similarity as an attribute. But when I look for example at the Result Window Layout, the Similarity is selected from a attribute list.

Wolfgang

[Mario]

When you right-click a group bar you have a context menu with additional options (e.g. collapse all).
You cannot filter on similarity, such a filter does not exist. Similarity is a volatile value which only exists temporarily in a result set.