computer vision - Possible to use a deep learning network on a collection of still images? -


i have collection of thousand trading cards different pictures on them. have database of high resolution scans of every 1 of these trading cards have ever been printed. i'd feed scanned images deep learning network, such if hold 1 or more of cards in front of camera, able identify one(s) holding.

it looks jetpac might place me start. have experience machine learning, numerical analysis, not image processing. examples i've seen far show people filming thing they're interested in identifying, , being able identify it. but, able dump folder of images in training data instead?

lastly, i'm aiming implement on system of raspberry pi 2's i've networked work in parallel. i'm not sure jetpac explicitly able support distributed computing, figure may able split video feed multiple feeds, , run each feed separate instance of jetpac on separate rpi.

am thinking problem right way? different approach more practical? help!!

edit: fear of sounding question general, question whether jetpac (or other deep learning library) able given collection of still images , trained pick instances of images out of video source.

you use deep-learning task, it's not advised unless you're dealing specific cases. task, want find specific image in image, , not learn set of images abstract class (like how dog looks like) done in learning tasks.

there plenty of known, fast , accurate image processing algorithms deal problem, based on feature extraction original images, , feature detection in unseen image. can see specific implementation of solution using opencv here, or learn more reading on feature detection.


Comments

Popular posts from this blog

c# - Binding a comma separated list to a List<int> in asp.net web api -

Delphi 7 and decode UTF-8 base64 -

html - Is there any way to exclude a single element from the style? (Bootstrap) -