Join GitHub today
GitHub is home to over 31 million developers working together to host and review code, manage projects, and build software together.Sign up
How to use dense sift in multi-class image classification #150
I intend to compare the classification results of small objects (64x64) at fixed size between HOG and dense sift. As I noticed that the number of feature descriptors from HOG is one, while the number of feature descriptors are many from dense SIFT. I am thinking of two approaches to in dense SIFT case. In the first case, I concatenate all the descriptors together into a single one. In the 2nd case, I use the bag of words approach, create a codebook first, then use the histogram of features as a final feature vector before sending it to a classifier.
Which approach do you think is more appropriate?