personalrobotics / linemod Goto Github PK

HOW TO USE LINEMOD

*Add object mesh:*

// here, plastic_plate is the name of the object, wahtever is its description (not used anywhere..)
rosrun object_recognition_core object_add.py -n "plastic_plate" -d "wahtever" --commit  

// once executed, the above command will return an object id, which is used in the next command for adding the mesh
rosrun object_recognition_core mesh_add.py <object_id> <object_mesh.obj> --commit
// eg. :
// rosrun object_recognition_core mesh_add.py 3f0abec62876fd72445b1d3da600739b ~/meshes/plastic_plate.obj --commit
// NOTE: the .obj and corresponding .mtl MUST have absolute paths to the .mtl and the .png (texture) respectively


*Run trainer on mesh*
rosrun object_recognition_core training -c `rospack find object_recognition_linemod`/conf/training.ork 

// the training.ork file is in herb2_ws/linemod/conf/ right now. 
// NOTE: currently, you cannot provide different training params for different objects, all use the same params.

*Run detection!*

rosrun object_recognition_core detection -c `rospack find object_recognition_linemod`/conf/detection.ros.ork

// again the detection.ros.ork file is in herb2_ws/linemod/conf/
// again, currently, you cannot provide different detection params for different objects.

// The detector publishes on /recognized_object_array
// To visualize on rviz, follow instructions from here: http://wg-perception.github.io/ork_tutorials/tutorial03/tutorial.html

// ISSUES:
// The osmesa renderer is not working on herb for some reason (gives garbage render output), so using GLUT renderer
// which is much slower. probably some gl linking problem with graphics card or someting..