Can someone point me in the right direction?

[mattqatsi]

I'm trying to make a small image comparison program for specific use.
It would compare images of a certain kind and would organize them with how similar.
It basically points out duplicates and similar cropped photos.
There is of course much more, but I just need a pointer to what to study
What should I look for to study?

[david blaine]

right direction: start by making less threads.

[se7enwords]

those "which celeb do you resemble most?" sites use that kinda shit...snoop in that direction?

[barbie]

i can point with my nis!

[mattqatsi]

Not like face recognition, but... pixel recognition.
Okay amit, I'll stop, but I need a better hint

[david blaine]

i can point with my nis!
-->

[se7enwords]

this is a lotta work just to organize ur pr0n bro

[mattqatsi]

this is a lotta work just to organize ur pr0n bro
Fuck, you realized what it was for

[Vershun]

Quick and dirty for crop detection:

Split up the image into something like 20x20 chunks. Go through every pixel in the chunk and get an average number for R, G, and B. Do the same for the other picture. Use the picture with the smallest number of chunks and do a Euclidean distance measure of the average RGB comparing the top-left chunk of the smaller picture with every chunk on the larger picture. Take the smallest Euclidean distances where location of chunk - width of smaller image and go through all the smaller picture's chunks with the corresponding chunks on the full image (using the offset). The smallest Euclidean value of all your potential crop places will be the most likely to be cropped.

I'd do trial and error to find a proper threshold. And the smaller the chunks you split up the better crop detection you'll have.

Just a thought.

Here's a textual penis _))======D

[mattqatsi]

Quick and dirty for crop detection:

Split up the image into something like 20x20 chunks. Go through every pixel in the chunk and get an average number for R, G, and B. Do the same for the other picture. Use the picture with the smallest number of chunks and do a Euclidean distance measure of the average RGB comparing the top-left chunk of the smaller picture with every chunk on the larger picture. Take the smallest Euclidean distances where location of chunk - width of smaller image and go through all the smaller picture's chunks with the corresponding chunks on the full image (using the offset). The smallest Euclidean value of all your potential crop places will be the most likely to be cropped.

I'd do trial and error to find a proper threshold. And the smaller the chunks you split up the better crop detection you'll have.

Just a thought.

Here's a textual penis _))======D
Thanks, but, uh, I mean like, what should I study?
What sort of language would be best to write it in?

[barbie]

Quick and dirty for crop detection:

Split up the image into something like 20x20 chunks. Go through every pixel in the chunk and get an average number for R, G, and B. Do the same for the other picture. Use the picture with the smallest number of chunks and do a Euclidean distance measure of the average RGB comparing the top-left chunk of the smaller picture with every chunk on the larger picture. Take the smallest Euclidean distances where location of chunk - width of smaller image and go through all the smaller picture's chunks with the corresponding chunks on the full image (using the offset). The smallest Euclidean value of all your potential crop places will be the most likely to be cropped.

I'd do trial and error to find a proper threshold. And the smaller the chunks you split up the better crop detection you'll have.

Just a thought.

Here's a textual penis _))======D

O_O

...

...

















































...

[Vershun]

Thanks, but, uh, I mean like, what should I study?
What sort of language would be best to write it in?

For image comparisons? It all depends on what your doing. Crop detection is easy like I posted, for harder stuff (fuzzy logic) things like neural networks and matrix transformations are applied.

So iunno neural networks and linear algebra?




PS notsober.

[mattqatsi]

For image comparisons? It all depends on what your doing. Crop detection is easy like I posted, for harder stuff (fuzzy logic) things like neural networks and matrix transformations are applied.

So iunno neural networks and linear algebra?




PS notsober.
Nor am I, thanks

[tispe]

basically dude your not gonna go from "Hello World" to fucking facial recognition in 2 steps idk wtf your asking itt

[david blaine]

basically dude your not gonna go from "Hello World" to fucking facial recognition in 2 steps idk wtf your asking itt

[mattqatsi]

I finished that Malik C++ book and have moved on into another game design book right now

[ganj]

riiight

[david blaine]

reading books !== talent

[mattqatsi]

reading books !== talent
reading books helps foster talent

[Foq]

You should spend a few weeks "lurking" the forum, watching what other people post, in order to get used to the general flow of things. During this time it is OK to respond to other's threads, but you should refrain from clicking the new topic button.

[tispe]

reading books helps foster talent
so i've read certin assigned chapters of books and i've taken classes and i've wrote programs and done projects for college classes for the past 3 or 4 years and I have yet to even come close to anything as complicated as what your thinking about doing

[RonMexico]

Yeah,

Planning Intricate Program -----> Learning How To Program


You are most certainly doing it in the wrong order.

[Mom]

dream big amirite

[Vershun]

OK well buzzed ranting aside the whole field of image analysis is fucking huge, you have to know exactly what you're looking for. I have very, very little experience with it (I wrote a thing that makes a mosaic automatically out of a bunch of little pictures and I went to a couple colloquiums on image processing for robotic path-finding). For copied and cropped images the thing I described should work fine. If you want to go into fuzzy categorical analysis you're going to need a good amount of AI/machine learning knowledge (kNN's, cNN's, neural networks, feature mapping, etc). For doing stuff like filters and image manipulation straight linear algebra and dynamic programming algorithms will work.

Point is you have to be a lot more specific on what you want to do before you can even think about what to start researching. Programming knowledge doesn't have anything to do with this stuff; a baby could learn it using C and imagemagick.

[mattqatsi]

so i've read certin assigned chapters of books and i've taken classes and i've wrote programs and done projects for college classes for the past 3 or 4 years and I have yet to even come close to anything as complicated as what your thinking about doing
Yeah, but you don't have the drive or ambition to actually attempt to tackle a project of this caliber. All my programming knowledge didn't come from class, I flunked out of that class...


Planning Intricate Program -----> Learning How To Program
I have something to aim for now. I can align my learning to take on that topic


OK well buzzed ranting aside the whole field of image analysis is fucking huge, you have to know exactly what you're looking for. I have very, very little experience with it (I wrote a thing that makes a mosaic automatically out of a bunch of little pictures and I went to a couple colloquiums on image processing for robotic path-finding). For copied and cropped images the thing I described should work fine. If you want to go into fuzzy categorical analysis you're going to need a good amount of AI/machine learning knowledge (kNN's, cNN's, neural networks, feature mapping, etc). For doing stuff like filters and image manipulation straight linear algebra and dynamic programming algorithms will work.

Point is you have to be a lot more specific on what you want to do before you can even think about what to start researching. Programming knowledge doesn't have anything to do with this stuff; a baby could learn it using C and imagemagick.
Thanks, I'm a little hungover to absorb and contemplate what this means, but it sounds insightful