Ask Your Question

Revision history [back]

Flann with large image database (>10000 images)

I am trying to build a image search system where I have a database of over 10000 images. I'm extracting AKAZE descriptors of all the images and taking 500 best points. I construct a large matrix of my image database and then pass it to Flann Index class. Then I am able to run a image search on my database.

The problem is, this method works fine if database is small than 2000 images with each image having < 500 keypoints. But when I use this with a larger database of over 3000 images, it won't give me proper matches.

PS : I'm using autotunedindex values.