Ask Your Question

Is it a reasonable expectation for OpenCV to be able to do object recognition on furniture?

asked 2013-10-16 20:03:03 -0500

jordan112 gravatar image

updated 2013-10-17 13:29:59 -0500

berak gravatar image

I'm building an application that allows you to upload lots of snapshots around your house, for example, doing an inventory of your living room: Couch, TV, chair, lamp, etc

I want to save this image, then have object recognition on it, so it can describe it.

Most of the examples I see are with facial recognition or basic shape recognition.

For example, a couch:


If so, how would I approach this? I think I would need to train it with a classifier, then go from there.

edit retag flag offensive close merge delete


in the meantime, read a bit

berak gravatar imageberak ( 2013-10-17 06:57:53 -0500 )edit

1 answer

Sort by ยป oldest newest most voted

answered 2013-10-17 06:20:13 -0500

I think your suggestion of training a classifier is indeed the way to go. What you need to do is the following and repeat this for each object class you want to detect.

  1. Collect enough training data: positive images in all possible situations you want to detect and negative images of all possible backgrounds. However, since you are making an app that needs to be versatile, and applicable in many situations, you will need heaps of training data, since this is basically detecting 'in the wild'.
  2. Annotate all your training data, selecting the objects in your input images, which is a very time consuming step.
  3. Use the data to train a classifier. For openCV the best way to go for now is using the train cascade classifier algorithm.
  4. Use that model to detect objects and make an inventory.

However, you will need quite some time for data collection and annotation. A good idea might be to take a look at the 'deformable parts model' approach by Pedro Felzenzswalb, which is the leading detector for objects on the Pascal Voc Challenge.

The challenge contains several models for objects, including benches, which could be reused for your purposes.

Interesting links:

edit flag offensive delete link more
Login/Signup to Answer

Question Tools


Asked: 2013-10-16 20:03:03 -0500

Seen: 987 times

Last updated: Oct 17 '13