Boost The World: Cat Detection

Cat faces

LBP | HAAR | HOG

VISIONARY_cats_LBP_HOG

 


Cat detection is not a trivial task, especially if you want to perform it on ARM devices. Before using the following cascades read carefully this page to get the best performance and to know the terms of usage.


*NEWS*: since June 2016 vision-ary project joined ARGO Vision, an innovative firm that excels in visual recognition. For inquiry about cascades and more, please contact ARGO Vision.

Full frontal (with partial profiles) cat’s faces, trained with:

  • approx. 9.500 positive samples (randomly sampled)
  • approx 0.9B of negative sub-regions containing outdoor and indoor samples (70%-30%)
  • Training size w=40 h=40 (the best one tested)

LBP: visionary.net_cat_cascade_web_LBP

  • Features set (approx.) : 67.600 features
  • Training time: ~3 days
  • TP: ~ 95.01% of positive training set
  • FN: ~ 04.99% of positive training set
  • FP: ~ 1e-006% of negative training set

HOG: visionary.net_cat_cascade_web_HOG

  • Features set (approx.) : 100 features
  • Training time:  ~3 days
  • TP: ~ 95.07% of positive training set
  • FN: ~ 04.93% of positive training set
  • FP: ~ 1e-006% of negative training set

Harder Examples

 


VISIONARY_cats_LBP_HOG_dogVISIONARY_cats_LBP_HOG_hard

6 comments

  1. Interesting exercise about visual recognition.
    The idea is good and applicable to others interesting use cases.
    WELL DONE ^_^

    • Thank you Roberta,
      you’re right: in the next days I’ll complete the pedestrians cascades (HOG, LBP) and in the next weeks I’ll share other cascades for visual recognition. The ultimate goal is handling a wide range of “in-the-wild” objects, directly on ARM devices.

  2. Hi,
    I’m interested in detecting cat and dog faces. Could I use your haar or lbp cascades or should I make my own ? Is that all that is necessary ?

    • Hello,

      the cascade is able to detect the cat’s face, you can build a binary cascade able to do the same about dogs. How can we help you?

      Vision-ary team

  3. Hi,
    Are the numbers you report based on the training dataset? Can you show your results on some standard pet recognition datasets like asirra or oxford-IIIt?
    -j

    • Hello John,
      this cascade is just for teaching purposes. No ambitions about products, there is no benchmark about the datasets you cited.

      Regards,
      Vision-ary team.

Lascia un commento

Il tuo indirizzo email non sarà pubblicato. I campi obbligatori sono contrassegnati *