Chris McCormick's profile - activity

Are you just trying to detect the presence of a car, or actually classify it's make and model? For make and model recognition, I've heard of people using SIFT.

I'm learning about the HOG & SVM human detector, and had a question about the SVM used. I've read that it's a linear SVM, and I think my question is basically why use an SVM if the classes are linearly separable (or at least roughly linearly separable)?

Here are my assumptions about what it means for an SVM to be a linear SVM, please correct me if I'm wrong about any of these.

For a linear SVM:

• No kernel function is used
• No support vectors are used in classifying an input
• The classification function is just a linear equation, with a weight applied to each component of the input vector.

Thanks for your help!