@cSKYmail2000 ... yes ...at 0.37 we can see a vertical white line outlining the model.a line appears along the crop line...black on left side of shoe and white on right side of the model at 1:28.
Thanks for your explanation, I understood your meaning. I keep the borderline just for next phase developing. In the other methods of depth map (ex. stereo), the borderline of forground could be mixed with background like Cambridge's I2I laboratory (using stereo camera). If removing the borderline is necessary on specific application, I can remove it through mixing with background.
I am working with Kinect for my Final Year Project. My working distance is 40cm~5m as you can see in my videos, can you show the 3D point cloud for a mid range configuration?
Sorry I haven't used the 3D library to generate 3D point cloud, is depth map OK? In my channel, you can watch it "Rough Edition of Depth Map Video in Medium Distance" or use the part of link: watch?v=8DkrXRRB82A
Yes, I can. What working distance do you apply to? I can adjust accuracy for different application. Ex.: It can apply in short distance for face recognition or medium distance like Kinect for recognizing human pose; long distance like surveillance system can also be applied but I still have no idea to apply for what problem in long distance.
I don't get the big deal... what is your background? To track a high contrast object against a white/friendly background doesn't even need your LED lighting... if you really want to show depth mapping, then show us also the real time video of what's behind your action figure, and make that a moving background. Then it will get interesting.
The clothes and fists of the doll are both "white" against the "white" wall; hair and face of doll are both "yellow" against "yellow" board. From 0:20 to 1:00, maybe you can check the 720p video again. Another question about motion background because I have the only one person; while I am moving the doll, nobody can be motion background at the same time. And computer vision is a lonely research, especially coding in midnight. In my channel, you can watch six demos about computer vision.
@cSKYmail2000 but you are not capturing depth... you are simply doing a feature extraction of an object with good contrast to the background, and your algorithm is probably well tuned for the dolls features. There is some application for this level of performance, however, but it's not a general solution to the requirement of foreground or background subtraction. Why do you need the LED illumination...? If you were capturing depth, then you could tell us where in 3D space various features are.
For your request, a new video "Motion Background & Motion Foreground" in my channel, that video is you said: "show us also the real time video of what's behind your action figure, and make that a moving background".
My friend help me to be the motion background.
The YouTube do not permit link address in the comment so I can't paste the full link in the comment. Following words is the part of link: watch?v=Kj7p8dBZZQo
This application doesn't need tracking function, it just based on depth map to analyze foreground information. So you can watch my hand on the screen. If you want to track multiple objects, basing the depth map will be easier than traditional computer vision. This patent is applying, publishing not yet.
Nice video. Can we remove the borderline around the character somehow?
observer699 1 month ago
@observer699
I cannot understand your meaning. Could you explain more about "character"?
cSKYmail2000 1 month ago
@cSKYmail2000 ... yes ...at 0.37 we can see a vertical white line outlining the model.a line appears along the crop line...black on left side of shoe and white on right side of the model at 1:28.
observer699 1 month ago
@observer699
Thanks for your explanation, I understood your meaning. I keep the borderline just for next phase developing. In the other methods of depth map (ex. stereo), the borderline of forground could be mixed with background like Cambridge's I2I laboratory (using stereo camera). If removing the borderline is necessary on specific application, I can remove it through mixing with background.
cSKYmail2000 3 weeks ago
I am working with Kinect for my Final Year Project. My working distance is 40cm~5m as you can see in my videos, can you show the 3D point cloud for a mid range configuration?
Thanks in advance.
MiguelAlgaba 3 months ago
@MiguelAlgaba
Sorry I haven't used the 3D library to generate 3D point cloud, is depth map OK? In my channel, you can watch it "Rough Edition of Depth Map Video in Medium Distance" or use the part of link: watch?v=8DkrXRRB82A
cSKYmail2000 3 months ago
Comment removed
MiguelAlgaba 3 months ago
Can you upload a video showing the depth map or a 3D point cloud with your camera?
MiguelAlgaba 3 months ago
@MiguelAlgaba
Yes, I can. What working distance do you apply to? I can adjust accuracy for different application. Ex.: It can apply in short distance for face recognition or medium distance like Kinect for recognizing human pose; long distance like surveillance system can also be applied but I still have no idea to apply for what problem in long distance.
cSKYmail2000 3 months ago
I don't get the big deal... what is your background? To track a high contrast object against a white/friendly background doesn't even need your LED lighting... if you really want to show depth mapping, then show us also the real time video of what's behind your action figure, and make that a moving background. Then it will get interesting.
tzotzo 3 months ago
@tzotzo
The clothes and fists of the doll are both "white" against the "white" wall; hair and face of doll are both "yellow" against "yellow" board. From 0:20 to 1:00, maybe you can check the 720p video again. Another question about motion background because I have the only one person; while I am moving the doll, nobody can be motion background at the same time. And computer vision is a lonely research, especially coding in midnight. In my channel, you can watch six demos about computer vision.
cSKYmail2000 3 months ago
@cSKYmail2000 but you are not capturing depth... you are simply doing a feature extraction of an object with good contrast to the background, and your algorithm is probably well tuned for the dolls features. There is some application for this level of performance, however, but it's not a general solution to the requirement of foreground or background subtraction. Why do you need the LED illumination...? If you were capturing depth, then you could tell us where in 3D space various features are.
tzotzo 3 months ago
@tzotzo
For your request, a new video "Motion Background & Motion Foreground" in my channel, that video is you said: "show us also the real time video of what's behind your action figure, and make that a moving background".
My friend help me to be the motion background.
The YouTube do not permit link address in the comment so I can't paste the full link in the comment. Following words is the part of link: watch?v=Kj7p8dBZZQo
cSKYmail2000 3 months ago
Can it track multiple objects, or is it just registering the IR reflectivity of the nearest object? Is the patent available to read?
madsciencetist 3 months ago
@madsciencetist
This application doesn't need tracking function, it just based on depth map to analyze foreground information. So you can watch my hand on the screen. If you want to track multiple objects, basing the depth map will be easier than traditional computer vision. This patent is applying, publishing not yet.
cSKYmail2000 3 months ago
Is the depth adjustable with software? Does changing the depth require changing the hardware?
osmaneralp 3 months ago
@osmaneralp
Yes, it can adjustable with software because it can capture depth map. Hardware does not require to be changed.
cSKYmail2000 3 months ago
very cool !!!!
jccordeiro 3 months ago