Does tod_training in object_recognition require fiducial markers on training images?

asked 2011-03-07 01:14:42 -0600

Julius
955 ●31 ●45 ●56

I have been following the Training Base Creation tutorial to create a database for recognizing objects using packages tod_training and tod_detecting in stack object_recognition. I noted that the supplied images for training (snapshots from different angles) all show fiducial markers.

Ultimately, I want to train the data base using data supplied by our University lab that has been collected without fiducial markers.

Is there a possibility to use tod_training without fiducial markers?

Comments

Did you every tried to make a training base with you own bag files? And did you succeeded? If yes I need your help..

dlmypr ( 2011-03-11 21:28:58 -0600 )edit

I have my own bag files with point clouds and 2d image data. I need to collect some ideas on how to work around this obstacle this week, but haven't done anything in this direction yet. I will let you know if I come up with something, and would be happy if you did so, too.

Julius ( 2011-03-13 21:47:40 -0600 )edit

Hey dlmypr make sure to also check out http://answers.ros.org/question/505/purpose-of-pose-estimations-in-tod_detecting and http://answers.ros.org/question/504/physical-setup-for-collecting-tod_training-data too. I think Alexander Shishkov has done a pretty awesome job on answering questions on tod_* today. If you're doing the training process please share your experience.

Julius ( 2011-03-23 09:40:24 -0600 )edit

add a comment

answered 2011-03-23 04:32:23 -0600

Alexander Shishkov
191 ●5 http://itseez.com/

updated 2011-03-23 08:27:17 -0600

I think it is impossible now. But you can modify tod_training (turn off finding fiducial markers and detection of transformation between current image and the some abstract coordinate system) :-) In tod_training we require fiducial marker(or markers) for detection transformation between different frames in bag file. For each frame we calculate transformation between the camera coordinate system and the some abstract coordinate system (where coordinates of chessboard corners are the same as in fiducial.yaml). Calculated pose is saved to Features3d->Camera->Pose. Using this information we can merge point clouds from different images to the one for example.

You can use other fiducial markers (one or some chessboards or grids of circles). Parameters of your fiducial marker/markers should be in fiducial.yaml file. tod_training package contains tool for generating this file.

edit flag offensive delete link

Comments

Also I recommend you to learn old version of tutorial (for example http://www.ros.org/wiki/tod_training/Tutorials/BaseCreation?action=recall&rev=17). They contain some useful information about fiducial markers and poses.

Alexander Shishkov ( 2011-03-23 05:04:57 -0600 )edit

Why is the current wiki less elaborate?

KoenBuys ( 2011-03-24 00:47:34 -0600 )edit

I don't know. Ethan supports this wiki page. I like more the earlier version.

Alexander Shishkov ( 2011-03-24 00:54:01 -0600 )edit

I like the high-level info about the fiducial markers on rev 17. But doesn't switching off the transformation leave us without pose information that is required later in tod_detecting when trying to recognize objects?

Julius ( 2011-03-25 06:27:42 -0600 )edit

add a comment

Does tod_training in object_recognition require fiducial markers on training images?

Comments

1 Answer

Comments

Question Tools

Stats

Related questions

Does tod_training in object_recognition require fiducial markers on training images? edit

Comments

1 Answer

Comments

Question Tools

Stats

Related questions

Does tod_training in object_recognition require fiducial markers on training images?