Closed miguelriemoliveira closed 4 years ago
Data from @tiagomfmadeira
Hausdorff Distance computed Sampled 2831151 pts (rng: 0) on MeetingRoom4_original_ICP.ply searched closest on FARO_meeting_room_vox_clipped.ply
min : 0.000031 max 0.050080 mean : 0.018958 RMS : 0.022740 Values w.r.t. BBox Diag (20.692211) min : 0.000002 max 0.002420 mean : 0.000916 RMS : 0.001099
Hausdorff Distance computed Sampled 2929856 pts (rng: 0) on MeetingRoom4_optimized_ICP.ply searched closest on FARO_meeting_room_vox_clipped.ply
min : 0.000050 max 0.049970 mean : 0.017471 RMS : 0.021115 Values w.r.t. BBox Diag (20.649876) min : 0.000002 max 0.002420 mean : 0.000846 RMS : 0.001023
Original on top, optimized on the bottom.
Hi @tiagomfmadeira , so I guess the optimization works. The Hausdorff goes from mean 0.018 to 0.017, so, an average gain of one millimeter. Not significant I would say.
The first image is clearly better on the corner for the bottom image, thus the optimized improves.
Do you agree with these analyses?
Simple suggestion: set the meshlab background color to white. That way, all these images can go directly to your thesis.
@miguelriemoliveira I was thinking of using these... but now that you mention it, a white background would indeed make it look better... Also should I change the color map too??
Yes, a white background is better, especially for thesis and papers ... and we want a lot of those :)
Which colormap you are referring to? the one of the points clouds for each camera? Yes, it s not adequate. It should be one like lines or something from the qualitative colormaps, such as the paired from here
Yes the one that maps each point cloud to a different color... I'll have to look into that
Issue, create and Issue before doing anything else :)
Issue, create and Issue before doing anything else :)
Discussed further in issue #31
Hi @tiagomfmadeira , I was housecleaning and found this issue, these images were quite nice ...
After having found two capital errors (see #25 and #22 ), we need to reevaluate accuracy of the aruco based optimizations