CODEWITHSUNDEEP
matlabmachine-learningsvmfeature-extraction
ytm
ytm

Reputation: 503

SVM training of HOG descriptor results (in Matlab)

I wish to classify cars by extracting hog features of positive and negative training samples. The problem is that I'm not sure what to do with the HOG features I acquired from each image in order to "convert" them into trainable data vectors.

Edit: Thanks, that clears out some things. I was already trying to concatenate the matrix as Bentoy13 suggested (thanks) but was unsure about what dimension to concatenate. I just have one last question, using this method means I have to re-scale all my training images to the same size. So I was wondering if that will still enable reliable classification. If it doesn't, how can I overcome this problem?

For others who may have questions about the process of extracting hog features, I just found this tutorial which is very helpful in understanding the HOG descriptor and its uses.

Upvotes: 2

Views: 1206

Answers (1)

lennon310
lennon310

Reputation: 12689

use reshape(h,[],1); or directly h(:) to vectorize the histograms inside the block. You may consider normalization for each vector as well.

Upvotes: 2

Related Questions