Using implementation of sift by A Vedaldi

first of all i should say i'm not very good in english so i hope you can get my point! does a开发者_高级运维ny body works with A Vedaldi implementation of sift? my final project in university is about object recognition and i have to work with sift... i use A Vedaldi implementation and i can find features and descriptors for similar images...but i don't know how to find best features from for example 10 images and build a data base for a special object to be recognized later... plz help me :(

I have worked with Vedaldi's SIFT. I used to choose SIFT features at the edges or better at the corners of the image. I find edges with canny, and corners with Harris detector. Then I keep points with a high score in corner/edges. When you use Vedaldi's SIFT you can pass to the function the points where you want to calculate descriptors. I pass corners points.