struggleforbetter

I'm also puzzled with this problem.The jaccobian should be the measurement change with the bias change. Why is it this? Are the two Jaccobians same ?

I also encountered this problem.on my computer, not using GPU acceleration, but faster.I track the image both with and not with GPU acceleration simultaneously.this is code `map FeatureTracker::trackImage(double _cur_time, const...

Turning on high performance on TX2 can achieve real-time, but I compared the GoodFeaturesToTrack of CPU and GPU and found that the feature points extracted by CPU are generally more...

it is easy to drift with intel realsense D435i,Sometimes, I put the camera on the desktop, the program will use the latest second frame as a key-frame, optimize td(I open...

launch: