Rotation invarient object recognition

Hi!

I recognize an object via HOG features. But it doesn't detect image's rotationals. Is there any way continuing at my point?

Secondly, in video, i detect with snapshot function. Maybe you know more efficient way..

Edit: get frame via getsnapshot(), than detect via step(dedector, photo)

frame= getsnapshot(handles.vid); bboxes = step(handles.Detector, frame);

Answer 1

[Image Analyst]

I don't understand your first question.

If you're detecting things in video frames, you do need to call getsnapshot(). That in itself doesn't detect anything - it just gets the frame so you can do the detection with an algorithm of your choice.

[mustafa]

You right. I use getsnapshot(), then detect it like photo. But getdata() or maybe other functions is more effective than getsnapshot()?

[Image Analyst]

Perhaps if you want to get a multitude of frames at one time instead of just one frame at a time. You still need to process just one frame at a time though, I think. Though if you're doing something like motion prediction you may have to look at several frames to predict the location in the next frame.

[mustafa]

Actually speed is not essential for my problem. Main problem is rotationals. Thanks for your support @ImageAnalyst. Sorry for my grammar :(