I am training a classifier of my own, specifically for detecting body arms. According to the training process I have to provide positive samples and negative samples, wich will be used in the different stages of the cascade classifier. As Mathworks says here, there is a tradeoff between having a higher false positive rate(wich I think is good) and the training set size you have, so, if I want a high false positive rate, then I have to provide a lot of positive and negative samples. I don't expect to provide picture by picture and selecting ROIs(Region of Interest) hundreds of times manually, so my questions are:

  • Is there any good repository of sample images for CV and specifically for body parts(I need arms specifically) different from this.?

  • Am I doing the right thing, or creating samples of my own is a better way to do this?

PD: Remember: I am interested in having a high false positive rate

  • $\begingroup$ False positive means the classifier tells you there's an arm even though there isn't one. Why would you want that? $\endgroup$ Commented Mar 30, 2013 at 8:42
  • 1
    $\begingroup$ @nikie: To work well, each stage in the cascade must have a low false negative rate. If a stage incorrectly labels an object as negative, the classification stops, and there is no way to correct the mistake. However, each stage may have a high false positive rate. Even if it incorrectly labels a nonobject as positive, the mistake can be corrected by subsequent stages. $\endgroup$ Commented Mar 30, 2013 at 15:05


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.