ROS Resources: Documentation | Support | Discussion Forum | Index | Service Status | ros @ Robotics Stack Exchange |
2012-09-07 21:21:03 -0500 | received badge | ● Famous Question (source) |
2011-12-27 05:42:17 -0500 | received badge | ● Notable Question (source) |
2011-09-13 18:49:06 -0500 | commented answer | Monocular Vslam I too see only the first frame. No points in the point cloud too. |
2011-09-13 18:45:02 -0500 | received badge | ● Supporter (source) |
2011-09-11 20:33:59 -0500 | received badge | ● Popular Question (source) |
2011-09-10 23:28:02 -0500 | commented answer | Monocular Vslam I tried it off the integrated camera on my laptop. The motion consisted of bumpy translation along a corridor - I was carrying the laptop by hand. I recall that the usb_cam node was not publishing proper camera info. Could this be a problem? |
2011-09-10 23:21:46 -0500 | commented answer | Monocular Vslam Regarding the assertion error: I too was initially getting this assertion. The nested use of "==" is the cause. Just replace the assertion with assert(frame1.pts.size() == frame1.goodPts.size()); assert(frame1.goodPts.size() == frame1.kpts.size()); |
2011-09-02 12:19:09 -0500 | marked best answer | Monocular Vslam vslam is unsupported experimental research code, and more specifically, the monocular pose estimation only works for specific types of camera motions (i.e., will not work on a dataset with unconstrained motion). Unfortunately, there are no current plans to improve the functionality of monocular vslam. However, patches are welcome. |
2011-08-31 22:13:49 -0500 | received badge | ● Student (source) |
2011-08-21 22:42:12 -0500 | received badge | ● Editor (source) |
2011-05-05 20:32:32 -0500 | asked a question | Monocular Vslam I am unable to get vslam_system 's run_mono to work on the 3VD & Versailles Rond datasets. It says "Returning a small amount of good points: 0, minGoodPts = 30". Also after running SBA the error is of the order 10^3. "Running SFM What could be going wrong? Has anyone managed to run "run_mono" on any published dataset? Sorry, I did not post additional detail last time. [ INFO] [1314008351.193699104]: Added frame to Mono Vslam system called PoseEstimator2d::estimate for frames 1 and 2 Number of points: 92, number of matches: 100 extractPnPData: frame1.pts.size() = 92 extractPnPData: good points 0 Running SFM Final error: 33044.761933 Dumping SFM returned r 3.128635 -0.279981 -0.019429 Dumping SFM return T 267.161316 3770.465820 9.712806 The number of epipolar inliers: 53 53 points before filtering 35 points left after filtering sba got 35 points Node projection matrix 1.2 0 0 -0 0 1.2 0 -0 0 0 1 -0 Node projection matrix 1.18084 -0.213058 -0.0148276 320.594 -0.213064 -1.18093 0.000851121 4524.56 -0.012286 0.00149597 -0.999923 9.71281 Added 35 points, 35 tracks sba.tracks.size = 35 costProj done sba.tracks.size = 35 costProj done Errors after sba initialization: 168.142901 896.277610 [SBAsys] Cameras: 2 Points: 35 Projections: 70 Hessian elements: 105 [SBAsys] Average 2 projections per track and 35 projections per camera [SBAsys] Checking camera order in projections...ok [SBAsys] Number of invalid projections: 0 [SBAsys] Number of null tracks: 0 Number of single tracks: 0 Number of double tracks: 35 [SBAsys] Number of disconnected cameras: 0 [SBAsys] Number of cameras with <5 projs: 0 [SBAsys] Number of camera connections: 0 which is 0 connections per camera and 50 percent matrix fill [SBAsys] Min connection points: 1000000 Max connection points: 0 Average connection points: -nan [SBAsys] Total connection projections: 0 [SBAsys] Connections with less than 5 points: 0 10 pts: 0 20 pts: 0 50 pts: 0 sba pointer: 0xbfc1d35c Removed 22 projections > 2 pixels ...(more) |