CBIR, Categorization, Patterns, Descriptors, Image recognition, SFIR, Computer Vision ... do you know what these words mean ? Ok, keep reading ;)
Use the technology you want. I will describe the project with a simple example. All should be run in Linux, to work through console commands.
1) need a script to insert tagged images. So the script just has 2 inputs: "image file path", "tag"
Then, imagine I insert in the system about 1000 images (I guess there will be some kind of database background). The images can be anything (a bottle, a car, a face, a brand logo, an elephant, a book cover, a painting, ... whatever!). And I have them tagged (one image = one tag). Imagine 7 of those 1000 images are elephants (as the ones you can find here: [url removed, login to view] )
This script has not to be very fast. I mean, it's not a must.
2) I need other script with just one input: an image file path. The output should be the tag of the most similar image of the database. For example, if this image to recognize is an elephant (obviously it's not one of the 7 elephant images inserted before, but a similar one), the script should output "elephant" (the tag)
* I understand that the more distint elephant images I inserted with the first script, the more easier and accurate the second script will say it's an elephant, since the "training" script (the first one) will save features, colors, shapes, and information about.
** This script has to be very quick. Let's say we have a group of about 100.000 images inserted, it should take a few seconds as much.
16 freelancere byder i gennemsnit $502 på dette job
Hello. Please see my profile. This is what I do. The price is high because I guess I will have to do a lot of testing etc., but I will assist you till we have good results.