Ask Your Question

Ahmad Sanny's profile - activity

2016-01-18 08:52:28 -0600 received badge  Nice Question (source)
2014-04-28 02:45:13 -0600 commented question The sign in CvSVM::predict

thank you so much, @StevenPuttemans.

2014-04-26 02:47:40 -0600 commented question The sign in CvSVM::predict

thanks @StevenPuttemans. your help is most appreciated. just pardon me for my last question, does the negative response in my experiment is due to training the positive set first then the negative set?

2014-04-25 10:24:21 -0600 commented question The sign in CvSVM::predict

@StevenPuttemans Hi there, do you mean -1 for positive and 1 for negative or it doesn't matter? thanks

2014-04-25 06:28:32 -0600 received badge  Editor (source)
2014-04-25 06:22:39 -0600 asked a question The sign in CvSVM::predict

Hi all,,

I'm using a binary SVM to classify positive and negative sets of images encoded using HOG descriptors to the machine; I labeled my positive set '1' and '0' for the negative one during training.

Based on the OpenCV manual, using float response = predict(inputMat, true); returns the signed decision function value. Does the negative sign denotes a Positive label '1'? I made few experiments* to reach this conclusion and I just want a confirmation from anyone knowledgeable/experienced with SVMs.

*The experiments was applied on 2 positives and 2 negatives using float response = predict(inputMat, true); once and then using float response = predict(inputMat); //Default false, returns label. I hope my conclusion applies to any image. Thank you :)

2014-04-04 05:47:22 -0600 asked a question HOG.compute input window dimension

Hi all. Apparently the .compute in HOGDescriptor is limited to specific window sizes. Can anyone help me what window sizes are supported? I couldn't find any documentation for this function. Say I want to compute the HOG feature for a training image of 64x40 (width x height). Thanks

2014-03-17 09:08:24 -0600 received badge  Student (source)
2014-03-17 08:51:20 -0600 asked a question Number of training images for HOG people detection

Hi all. Can any one give his/her experience on training SVM for people detection using HOG? I don't mean how to do so, but I mean how many images for training is sufficient for successful detection: How many positive and so negative images.

If we consider INRIA person set that was used in Dalal-Triggs' paper, I went through the folders and files and I got confused with the following:

The folders are organised in a way that seems incorrect for me, or perhaps I understand the term 'train' and 'test' wrongly. There is a folder named '70X134H96' and contains a subfolder 'test' then 'pos'. Is this folder meant to be positive images for training? If no, then which folder?

How is it to train negative images? Do I crop randomly from the 'neg' folder and load it into the SVM as a negative image?

Thank you :)

2014-03-03 06:05:25 -0600 received badge  Supporter (source)