Project 2 Artifact Page
CSE 455: Computer Vision
Project 2 Artifact
Lai Kwan Wong and Rowan W. Hale
Test sequence
Sequence with Kaiden panorama head
Taking the panoramic shot was pretty simple, though there was concern that our shadows would end up looking odd or messing up the panoramic stitching.
In the end, that wasn't a problem and it turned out just fine.
Sequence taken by hand
Taking the shots by hand was a bit difficult, because it was hard to keep the camera stationary while moving around it.
The camera ended up tilting downwards over the course of our shots, which is part of the reason that the stitching is blurred in places.
ROC comparison of our code vs. SIFT
Here we provide ROC plots for both the Yosemite and graf images.
This plot shows that the success rate of SIFT is significantly higher than MOPS.
The lack on any MOPS data in this plot is interesting, but we still see that SIFT is performing well.
Here we show the output of the Harris operator for Yosemite1.jpg and img1.jpg (of the images/graf/ directory).
