501 - not implemented's profile - activity

hello, In openCV I use binary descriptors like ORB and A-KAZE. Now I want to test and perform some matching algorithms, that aren't part of opencv. For this, I want to create a BitString of each row of the descriptor. But there I have some issueas. How are the bits saved in the the mat object?

thank you

ah...of course...but no, I want not use intern functions of opencv for this. I want only extract the string representation

no I doesn't normalize the descriptor. Is this important? And after the normaliztion, I can easy extract each row and concatendate the byte to a integer?

Hello, I'm working with binary descriptors like ORB and AKAZE in opencv and want to store each descriptor (a col of the descriptor mat) as binary string. To perform the query, I want to save these representation as a array of integers (int = 32 bits and I have a 32 bit machine), so I can easy compare binary string with hamming distance.

My problem is to extract the bit representation from the mat because the at has byte representation. Is there something that I must conside if want to extract the bits?

thank you! :)

okay thank you. Any other ideas or solutions? I also can save my descriptors as a binary string

Hi, I thought that I can use geoNearfor radius serch (https://docs.mongodb.org/manual/refer...)

Hello, I'm looking for a database to store my extracted (surf, sift, orb,...) features.

My idea is to use MongoDB as feature store and add each row of image descriptor as document to MongoDB.

MongoDB also includes knn and radius serch algorithms, but I can't estimate how it performs on image features

(curretly I have descriptors with 61 dimensions and more than 10.000.000 features).

Has someone ideas or experiance with mongodb or another socument store?

thank you and best regards

okay sorry, yes the DescriptorMatcher has read and write methods but I think it only used to load the properties for the matcher (algorithm,...)

Can I use JavaCV to save a trained index?

yes, i mean the DescriptorMatcher. Is there a work around?

Hello,

Can someone help me to save a trained flann index?

In time I serialize all descriptors to json and reoad all descriptors after the program is started.

For a lot of descriptors this takes a long time.

Is there another method to save the flann index with java?

yes, that's theI the aproach I'am going. At the moment I compare all images with each other. But it is very difficut to save the intersection from all good matches of all images. Is there a datastructure that can help me to solve this problem?

thanks!, that's a beginning for images, that shooted by a tripod. Any ideas how to handle keypoints that are near similar but shooted from a differet position?

I think no, bechause I must set up the count of clusters

Hello, in a project I'm using SURF to extract keypoints from a set of near similar images. The images are near similar, so they have a lot of similar keypoints and a few keypoints that are different. My goal is to find groups of keypoints that matched by all images

is there an idea how to solve this? I can't use kmeans, because i doesn't know something about the k

UPDATE

My current approach is using a bruteforce-matcherto find the good matches of each image. I compare all images with each other. But it is very difficut to save the intersection from all good matches of all images. Is there a datastructure that can help me to solve this problem?