search

AlexeyAB / darknet

YOLOv4 / Scaled-YOLOv4 / YOLO - Neural Networks for Object Detection (Windows and Linux version of Darknet )
http://pjreddie.com/darknet/
Other
21.8k stars 7.97k forks source link

Optical flow code #565

Open ahsan856jalal opened 6 years ago

ahsan856jalal commented 6 years ago

@AlexeyAB Can you please refer me the paper you followed for Optical flow

fvlntn commented 6 years ago

@AlexeyAB would like to know too

AlexeyAB commented 6 years ago

@ahsan856jalal @ralek67 I don't remember it ) It was a years ago. If we chose from approaches that are implemented and built-in OpenCV (except OpenCV-contrib):

Then the most accurate and the fastest Multi-object tracker is sparse Optical Flow (PyrLK on GPU)

There are another 5 different visual object tracking techniques in the OpenCV-contrib: https://github.com/opencv/opencv_contrib/tree/master/modules/tracking

If you know more better Multi-object tracker that already implemented and won in some competitions then let me know.

alexanderfrey commented 6 years ago

How is deepsort tracker performing in comparison to PyrLK ? Does anyone have experience here ? I'm using deepsort and it's performance is quite nice...

AlexeyAB commented 6 years ago

@alexanderfrey Hi, Do you use GPU implementation of Deep Sort? What repository do you use?

I think most of modern tracking techniques have better accuracy than PyrLK, but PyrLK is faster due to simple algorithm and implementation on GPU.