Geek asks and answers

Cat Detection Using Haar Training

I would like to detect deformable objects (CAT in my case) using HAAR cascades. I tested 2,000 positive and 3,000 negative images in 20 stages. I do not see anything expected. I am sure this may be the problem of my training dataset. My positive dataset contains images of a cat. In different forms and poses of the same cat. I wonder if Haar can recognize the differences in the positive dataset and still recognize them as cats.

Need help. Is there any other alternative method that I could use?

+5
source share
1 answer

Firstly, you did not indicate what you are using to train hara classifiers. If you are not using opencv_traincascades, this is the place to start.

Secondly, finding a cat is a very difficult task, there are many uncertain edges, rotational and spatial dispersions. using this train of thought, answer these questions to yourself:

  • What does your positive dataset look like?
  • This is similar to what you show through the input channel.
  • What is the level of false alarm of your classifier when you trained it?
  • What do you include in your background set (more images that look like cats that don’t like cats, etc.).

I would suggest publishing my conclusion in the question from the classification stage so that we can better understand what is happening. Take a look at these links for more help.

, , openCV ? (, , , )

G./

+4

More articles:


All Articles