Main Area and Open Discussion > General Software Discussion

image comparer desktop software

<< < (6/7) > >>

Contro:
now is finished 22.06.2013.17.32
in the left column the options to find
I will comment......
I was eating by the way..... need to finish first
 :P

Is general recognition. An interesting voyage to my images, but don't see powerfult searching options like search for a determined face.
You can obtain faces "in general".

 :-*

Edited 22.06.2013.18.41 :
I don't worry about the space for reindexing, but seems a program needing for powerful options.
I have read the comments about other similar program and beginning to think is not a easy question find a program with fullfill your requeriments
 :-*

sicknero:
After saying you may be out of luck, here's something I've found: Image Forensics Search System-4wd (June 22, 2013, 02:01 AM)
--- End quote ---

Apologies for cropping your post but I didn't want to fill the page. I got quite excited when I saw this so downloaded it and gave it a pretty thorough test drive today.

So, firstly it's pretty old. The website only mentions OSs up to XP and the installer looks old. It also requires JRE but, I just used Uniextract to extract the IFSS installer which includes a java folder and it seemed (at first - more on that shortly) to run fine like that without JRE on my PC. I'm guessing though that this also means that the included java version is pretty ancient.

Anyway. I randomly picked a folder of 103 photos of my local town, picked one out and created a selection of edits of it... I mirrored it, shrank it to a 75x56 thumbnail, grayscaled it, graffitied it, and rotated it through 90 degrees.

The program is easy enough to use, a nice simple layout and interface. You just load your initial pic, select a few options (some of which I still don't understand entirely...) then point it to the directory you want to search and let it go.

It's pretty slow. Took around 5 minutes to scan the 103 pics in the target directory, and while it's working javaw.exe is taking about 90-95% of cpu cycles which also means that the program itself is more or less unresponsive while working. I mostly had to use task manager if I wanted to cancel a search.

Results-wise through, it's great. It picked up on all the edits I'd made with the exception of the 90 degree rotation, though to be fair I think most duplicate finders fail on that too with the exception of Visipics.

It's different in its approach. When you select the target directory the program scans this and loads it into a window from which you can then select which pictures you want for comparison. Select all is the useful default, but this part of the process itself takes ages... two or three minutes for my 103 photos.

Then when it's finished, it doesn't show you a list of "hits" but lists all the files in the target directory, sorted hierarchically according to percentage match.

So far so good, but onto the main problem... I went on to try it on a folder of about 5,400 downloaded pictures, and five times it just exited itself after a few minutes for no apparent reason, before it had even loaded the folder. I even made a copy of the folder and renamed all the pics in it, thinking it might be an issue with path lengths (you know what long names downloaded pics can have...) but that didn't help. I then put aside my non-install version and ran the installer properly, but that made no difference.

So, mixed results really. I'm just running it now on a directory of 934 National Geographic photos and it seems to be fine with that (Edit - completed successfully). If I can find some definite cut-off point beyond which the program won't work, then I think it's definitely a keeper - depite its slowness and cpu-hogging java, it's still the only program of its type that I know of.

Interestingly, even when I ran the installer it still creates a java directory in the program folder rather than actually installing java. Not sure what to make of that.

Ultimately I guess, a better solution is if we could hunt down a duplicate finder that has the option to only show cross-directory matches, e.g. show when a pic in Folder A matches a pic in Folder B, but not when a pic in Folder B matches a pic in Folder B. I've been looking through various programs today but none of them seem to do this.

evamaria:
I'd been very intrigued by 4wd's find, and I want to say thanks for sharing - even if such an exotic program doesn't live up to our expectations, it's always a good thing to extend our knowledge beyond the classics. (I didn't find that prog by my searches, searching for 1 hour though, so it also seems to be a good idea to include the term "forensic" in your search whenever it might apply.)

Also, thank you very much, sicknero, for this big effort (that I, for instance, didn't have time to undertake) and sharing the results. Now you say, "Ultimately I guess, a better solution is if we could hunt down a duplicate finder that has the option to only show cross-directory matches, e.g. show when a pic in Folder A matches a pic in Folder B, but not when a pic in Folder B matches a pic in Folder B. I've been looking through various programs today but none of them seem to do this." - would you mind sharing the list?

(AntiTwin's got that feature "compare a folder against another", but of course, it's not specialised - I regularly even use it for photos, with 90 or 95 p.c. setting, with good results, but I've never tried to use it for specific tasks, like heavily cropped photos, etc., so I suppose it being a "general comparer", it will not be best here.)

And then, VisiPics perhaps gets too much prominence here? And Visual Similarity Duplicate Image Finder (40$, from MindGems) doesn't get enough prominence, presumably!

In fact, their screenshots are not very enticing, but people on the web say it's got the best algorithm, among those "photo comparing" programs - would you consider, sicknero, to trial it for us, with the same set of tasks you have used to trial the "forensic" thing? (Or is it impossible to do a "folder against folder compare" with it, even with the "prof" version?)

That MindGems thing has got a crippled "standard" edition, 25$, but without filters, and the regular 40$ edition, called "prof"... and then, there is a 500$ "corporate" edition, the sole difference being / seeming to be that it's "scriptable" or something, i.e. you can trigger it, with attributes, by command line.

Now that's a little bit ridiculous if we assume that the 40$ and the 500$ version probably share the same algorithm(s) - it would be interesting to know if there are 500$ programs that apply much BETTER algorithms than do their 50$ counterparts...

Anyway, if we assume they're not crooks, that'd mean their 500$ version was worth the price, after all... and that'd mean their 40$ version would be a definite "steal" - best buy of all those programs out there that we know (since they seem to indicate it's the same algorithms for 40, and for 500$...).

Hence my request to kindly trial this for us, could be a real discovery.

sicknero:
Oh good grief, how did I miss that..??? I guess I was stuck in a "need something new" mindset and didn't even see that option in AntiTwin. It does do what I was asking for and, it even has a "Search for a certain file" option, both of which work in the image search mode.

The results aren't great though - it picks up the grayscale image and the graffiti one but misses the mirror image and the thumbnail. The percentage selector only goes down to 60%.

The others that I have are Visipics which as you say is popular, but it does deserve it although it fails to pick up the grayscale for me. Duplicate Cleaner 3, Awesome Duplicate Photo Finder, AllDup, AntiTwin of course... some others too, I'm a bit of a software collector.
Actually just got a new one today, Similar Image Finder, which I'm quite impressed with so far, it's picked up all my test pics except for the mirrored one. It also has an option to ignore cross-directory matches, so I've been in touch with the developer today to see if they're open to feature requests and if it would be possible to add the other option.

But the Mindgems one, yes I'd love to take on a trial of it. I'll check it out properly when I get a chance, must first see if the trial download is functional enough to test it properly. Must admit I personally like the look of the screenshots, but then I don't much like the way that GUI design has been going for the last few years anyway.
I'd not considered payware, as the site where this initially came up is freeware only, and it's an interesting point about the difference in licenses. Reminds me of a cd burner that has a freeware version and then a commercial version that costs hundreds.

I'd want to spead more time on it probably... the way I tested the Forensics one was just based on what I'd personally like such a program to do, and also as you can see I've not properly looked at all the ones I already have! The point about cropped pictures is a good one too, I'd not thought of that and I think it definitely deserves including in a group of test pictures.

I'll post back in a few days anyway and let you know how it goes with the Mindgem one.

evamaria:
Great, and thank you very much!

For people who quickly want to check the components of this prog are about 25 years old:

http://www.mindgems.com/products/VS-Duplicate-Image-Finder/VSDIF-Screenshots.htm

(even more horrible than FastStone Image Viewer)

And the last image there is the filter window, and indeed, I don't see any "compare folder against folder" option there - it would be a pity if the lack of this functionality put aside the potentially very best prog of them all...

Of course, the "crop test" is far more important than any other, for "real use" - detecting "stolen" photos - but, oops, it just occured to me that we're speaking of photos here that are already on your hdd, and exclusively of these, so for detecting stolen photos within the web, you'd need google's photo search anyway... and THAT's perhaps the reason there is no advanced development, and no tremendously good 500$ tool out there: it's just not worthwile, for the aforementioned reason!

But this being said, a photographer who takes photos, of the same subjects, from wide-angle down to big zoom, could always be interested in having software to prefetch groups of the same subjects

- but would not be willing to spend 500$ for this, so the lesson here is, whenever you don't understand the market, think again. And yes, I'm able to imagine special police use of such software: comparing their "very special" photo collection (vice squad) to what YOU might have on your hdd - but there are special services for that in every country, so world-wide, this very special market would take about 300 licenses, make it 500, so we wouldn't speak of 500$ software here, but of 5,000$ software.

And as for our photographer, he wouldn't want to compare one photo with a set of others, most of the time, but would want to constitute groups of similar photos - again, the missing function "explains itself" by the scarcity of the demand for it?

But then, even for a wedding photographer, it'd be of interest to choose one person, on one photo, and then have automatically gathered all photos on which that person is present, even in a group, and this not only applies to the bride, but also to any person who might buy photos then (ok, they will browse anyway, but out of 1,000 photos, then? There will always be a pre-selection, of perhaps 150 "best shots", and then, 800 shots among which only SOME might be interesting to this person or another, so here software assistance might be of big help).

And then, why the need to put the "single" photo into a folder of its own, in order to have it "compared" to all the others? That's cumbersome! Why not a function "compare the current photo against all within a certain folder (and including its sub-folders)"?

All the more so since such functionality, coding-wise, and contrary to the implementation of better compare algorithms, is so easy a child could code it!

Ok, that's been written for some developers who might check google for mentions of their progs, and then read some remarks in order to get some ideas.

As for google pic search, if I were up to steal photos, I'd turn them by 180 degrees and heavily crop them, considering that light tonal fiddling would certainly not cause google to differentiate them, and heavy hampering with color, brightness and contrast would make the photo unusable anyway, so turning around then cropping seems to be the "best" policy - would be interesting if google detects such falsifications, especially, of course, if the cropping isn't centered.

And to finish this pêle-mêle post, google, some years ago, had photographed the streets of Germany. Then it appeared, inadvertently (!!! how do you do THAT???) they had, by this photographing buildings, recorded "sms" and other electronic messages sent by the tenants of these houses, too. And now, just these days, years later, German press discusses that these recordings have not yet been deleted (as they had promised once, though).

So much for applied forensics. ;-)

Navigation

[0] Message Index

[#] Next page

[*] Previous page