But very similar outputs it has. Orientation detection, robustness against illumination, noise, rotation, up to 30 degree view point change etc. it looks like SURF or a SIFT like feature set.
@dhisar Indeed, we estimate object orientation. But the SIFT algorithm estimates orientation for a single feature point only. Detecting full objects like cars robustly using a single feature point is impossible.
The algorithm indeed is robust to variations in light and image noise. All techniques for object detection are related and share basic algorithmic elements for recognition. However, our total approach is different and uses some proprietary elements.
We do not use the default OpenCV Haar-based detector in which you can import an XML file.
The detector shown in the video uses proprietary technology developed at ViNotion.
The detection quality is much higher than the default Haar-based detector.
Moreover, it is possible to extract the viewpoint of the vehicles during detection, even though no motion information is used (each frame is processed independently).
it sounds you hacked sift or derrived it...
dhisar 2 weeks ago
@dhisar This technique is not using SIFT feature points.
vinotionNL 2 weeks ago
But very similar outputs it has. Orientation detection, robustness against illumination, noise, rotation, up to 30 degree view point change etc. it looks like SURF or a SIFT like feature set.
dhisar 2 weeks ago
@dhisar Indeed, we estimate object orientation. But the SIFT algorithm estimates orientation for a single feature point only. Detecting full objects like cars robustly using a single feature point is impossible.
The algorithm indeed is robust to variations in light and image noise. All techniques for object detection are related and share basic algorithmic elements for recognition. However, our total approach is different and uses some proprietary elements.
Thanks for your input.
vinotionNL 2 weeks ago
Good job :D
your xml file is so better then my XDDD
amd5200killer 11 months ago
@amd5200killer
We do not use the default OpenCV Haar-based detector in which you can import an XML file.
The detector shown in the video uses proprietary technology developed at ViNotion.
The detection quality is much higher than the default Haar-based detector.
Moreover, it is possible to extract the viewpoint of the vehicles during detection, even though no motion information is used (each frame is processed independently).
vinotionNL 11 months ago
@vinotionNL
Good information for me,Thanks a lot :D
amd5200killer 11 months ago