Ask Your Question

how to classify a photoset

asked 2019-08-02 03:53:26 -0500

luca123 gravatar image

updated 2020-02-13 12:02:41 -0500

hello, guys, i want to recognize object with template matching in opencv, that means i need to store many template picture of object in a folder, and then use matching algorithmus to find the desired object from camera. but i need to label these template picture to define which belongs to which object. sothat i can get the name of the object with the matching algorithmus. i think about the databae to tag the picture, but i dont know how to do ? or it has any another ways to solve this? do you have any tip? thank you in advance.

edit retag flag offensive close merge delete



please do a course in computer-vision and machinelearning first. (this will answer all questions above, and sorry, it's not the purpose of this site to replace it)

berak gravatar imageberak ( 2019-08-02 04:18:40 -0500 )edit

1 answer

Sort by ยป oldest newest most voted

answered 2019-08-02 14:58:04 -0500

holger gravatar image

I recommend coursera - i am currently doing this course But you can also learn from the internet yourself. I still recommend this course as this topic is not trivial.

First of all - what is a classifier? A piece of software which puts the input data into different distinct categories (also called classes or labels or target values)

The thing is - you want your classifier to be generic . You most likely want to classify a car as a car regardless if its blue , green or black. Or having different height, amounts of doors etc. A way to archive this is to train on different training sample which should have these variations.

Template matching in opencv is the opposite of being generic. It will match exactly the input image but misses all the variations(height, width, blur, color, etc...).

So if you don't have these variations you are fine to use template matching for classification but its an unusual way to do it and most likely it will fail for your use case?!

So what to do: Learn how to use a neuronal network for image classification. There are tons of tutorials for different network "frameworks" (tensorflow, pytorch, yolo). Just pick one - if you are funny you can also use opencv to run them - but this goes to far right now.

edit flag offensive delete link more



In addition to the ML course by Andrew Ng, I would also recommend his DL courses on Coursera. These DL courses (particularly the course on CNNs) helped me get up to speed with the new techniques and trends in computer vision

phillity gravatar imagephillity ( 2019-08-03 17:14:16 -0500 )edit

Question Tools

1 follower


Asked: 2019-08-02 03:53:26 -0500

Seen: 274 times

Last updated: Aug 02 '19