ROS Resources: Documentation | Support | Discussion Forum | Index | Service Status | ros @ Robotics Stack Exchange
Ask Your Question

tod_detecting output

asked 2011-02-17 09:12:34 -0600

dlmypr gravatar image

Hi all,

I am using tod_training and tod_detecting packages for object detection and recognition. They work well but the output is just an image of estimated pose and the name of the recognized object.

My question is how can I find this estimated pose? Not just by seeing on the image.

edit retag flag offensive close merge delete


Hello, are you using tod on its own or via the Recognition Infrastructure (ReIn) ?
raphael favier gravatar image raphael favier  ( 2011-03-01 22:33:08 -0600 )edit
yes I am using it on its own. Actually I trained some objects from bag files with tod_training and than I detect those object from images or camera recordings using tod_detecting. I can get estimated pose visually on the input image I gave but I can't take it numerically. It computes it somewhere..
dlmypr gravatar image dlmypr  ( 2011-03-02 20:17:34 -0600 )edit
Ho, may I ask what did you compile it with? (Cturlte, Unstable, Diamondback?) I'm having a few compilation issues myself.
raphael favier gravatar image raphael favier  ( 2011-03-03 22:00:34 -0600 )edit
Diamondback. But debugged the code myself
dlmypr gravatar image dlmypr  ( 2011-03-14 21:17:03 -0600 )edit
Would it be possible for you to post a patch?
raphael favier gravatar image raphael favier  ( 2011-03-15 01:33:13 -0600 )edit

How you run tod_training and tod_detecting? Can you share instructions? I can't find tutorials on package doc. Thank you~

sam gravatar image sam  ( 2012-10-28 06:22:06 -0600 )edit

1 Answer

Sort by ยป oldest newest most voted

answered 2011-03-02 20:41:40 -0600

pbrook gravatar image

If you are running the recognizer node from tod_detecting and are seeing text output like

 "Object name = " << object.objectName << ", imageId = " << object.imgIdx

then that is generated from line 167 of recognizer.cpp.

Each of the ObjectInfo objects referenced there have rvec and tvec members (see this) which presumably store the pose of the object. You could modify that loop to print out or otherwise output the pose as well as the name.

edit flag offensive delete link more


Is tvec the translation vector camera frame to object frame and the rvec is the rotation vector from from object coordinate frame to camera coordinate frame? And what rvec exactly shows? radians? And angle changes with respect to which axises?
dlmypr gravatar image dlmypr  ( 2011-03-02 23:38:24 -0600 )edit

Question Tools

1 follower


Asked: 2011-02-17 09:12:34 -0600

Seen: 474 times

Last updated: Mar 02 '11