Open DonaldTsang opened 6 years ago
Hi @DonaldTsang . This was a very short-term POC project for a class, and the code is terrible. To answer your questions,
1) We did use k-means clustering, and we determined that there was 9 clusters 2) Yes. We used SIFT because that was the most popularly used method. The main idea behind this was Bag-of-Features, and you can switch out SIFT with other feature extraction methods. 3) The database stores the image index, its corresponding SIFT features, and its cluster assignment.