View Full Version : [Question(s)] Open source machine vision code to locate a human in a room?

05-05-2012, 09:07 AM
Has anybody seen a resource covering specifically the use of machine vision for locating a human in a room, including cases where they may be partially obscured? I know about OpenCV and other machine vision packages, but I'm looking for a paper/library/code example that focuses solidly on identifying a human in a room; something that is readily integrated into code rather than being a rough idea that would require weeks or months of tweaking or enhancing to do a reasonable job of succeeding at the task. I'd prefer something that does not require binocular vision and can work with a single camera.

-- roschler

05-05-2012, 06:49 PM
HOG (Histogram of Oriented Gradients) is pretty much the standard in detecting people in images -- but it really doesn't work with partially obscured images. There is a trained implementation in OpenCV, and a number of papers on the original algorithm and variations thereof.


05-06-2012, 11:33 AM
Thanks Fergs, I'll research HOG.

-- roschler