Fred Mili's profile - activity

Can you give me an example of how to "get rid of outliers early"? I think outlier removal is after matching step and not in detection phase. As you see, ORB constructor already has a value of 50 which means it only consider the 50 best keypoints.

I was playing with OpenCV Mixed Processing tutorial and I experienced a good framerate (~27) when detecting features in camera frames with FAST detector.

I changed the detector to ORB and the framerate dropped to around 10. Is this because ORB is not as fast as FAST or I'm missing something? Is there any way to tune up ORB and speed this up?

Mat& mGr  = *(Mat*)addrGray;
vector<KeyPoint> v;

OrbFeatureDetector detector(50); 
// FastFeatureDetector detector(50);
detector.detect(mGr, v);

@StevenPuttemans, thanks a lot. But one question: Is it possible to write an ORB descriptor to XML or YAML file? What would you do if you had to save an ORB Descriptor to a file and read it later on?

@StevenPuttemans, I didn't do that. Somebody else has done it.

@berak, it's just an ORB descriptor written to a binary file.

I have a binary file which includes ORB descriptor. I need to read it to a cv::Mat so I can use it for further detection. FileStorage only handles XML and YAML file. How can I read the file?