vigneshr95's profile - activity

I am using dense trajectories which tracks feature points in multiple spatial scales.

I need to cluster the extracted features using k-means.. should i do it as a whole or for each spatial scale?

Thanks a lot .. but this would require lot of time to complete for a single frame.. any other idea?

I have 2 vector of keypoints like this: vector<keypoint> points; vector<keypoint> keypoints; fea.detect(t,keypoints); goodFeaturesToTrack(t,points,100,0.001,8,CV_8UC1,5,false,0.04);

I need to store those keypoints which are at the same pixel location in both vectors. How can i do it? (eg., if there is a feature point at(3,4) in both vectors i should take it.. else i should not)

That is my doubt @stevanputterman .. My Annotation file is in .txt format .. how can i read that into my program and compare?

ReDID dataset .. Ya Since i need to use only the decriptors inside ROI for the next step.. I need to select those inside the envelope defined in annotation.

I have extracted features for the feature points in the entire frame. How to select the features in ROI using annotations? (Annotations of ROI are available in the dataset.)