Extract the human mask from the image

asked 2018-01-12 03:55:24 -0500

vps gravatar image

updated 2018-01-18 04:20:36 -0500

Hi,

I am working skeletonisation. In preprocessing, I have to find the human mask from the RGB images by using cnn or any other machine learning techniques. I want to extract only person area. C:\fakepath\Unbenannt.PNGThe mask should be sharper. It will be great , if you suggest any documentation or techniques. (I tried K means clustering, superpixel grouping , gradient techniques but unable to extract accurate mask. )

thanks.

edit retag flag offensive close merge delete

Comments

2

o_o that is actually quite accurately. Maybe you can have a look at how OpenPose framework solves the issue?

StevenPuttemans gravatar imageStevenPuttemans ( 2018-01-12 06:41:33 -0500 )edit
2

just curious, what is used for the skeleton here ?

berak gravatar imageberak ( 2018-01-12 07:04:29 -0500 )edit

Hi Steven...Thanks for reply. I already checked , they are directly finding the human joints. Here first, I want to find the human silhouette or mask from the image. Then, I will operate on this mask.

vps gravatar imagevps ( 2018-01-12 07:43:10 -0500 )edit

Hi Berak..I am using the RGB as well as depth data for the estimation of the joints but the depth data is not accurate.

vps gravatar imagevps ( 2018-01-12 07:45:21 -0500 )edit
1

Is there any reason why you want to do it differently than the current state-of-the-art open frameworks?

StevenPuttemans gravatar imageStevenPuttemans ( 2018-01-12 08:07:24 -0500 )edit
1

Hi Steven, I want to use different approaches :). Also, I am using superpixels with mean value to reduce the computational power.

vps gravatar imagevps ( 2018-01-12 08:17:43 -0500 )edit