object_recognition_clusters

Package to contain nodes and modules for publishing and processing clusters within the Object Recognition Kitchen.

###Modules

##Ecto

• io_clusters.PointCloudMsgAssembler that generates pointclouds based on a set of clusters
• ObjectClustersPublisher ecto plasm which connects to the tabletop pipeline with the PointCloudAssembler and publishes the

##Python

• object_recognition_clusters.ClusterBoundingBoxFinder which computes the pose of an object given it’s point cluster using PCA

###Nodes

• clusters_to_pose that runs the clustering the python module, used mainly for testing and demonstration.

Changelog for package object_recognition_clusters

0.1.0 (2014-06-23)

• Adding minimum version of sensor_msgs package
• Commented out debugging TF publishing. Was giving a continuous warning looking like: Unable to transform object from frame 'object_frame' to planning frame '/odom' (Lookup would require extrapolation into the past. Requested time 82.933000000 but the earliest data is at time 96.737000000, when looking up transform from frame [object_frame] to frame [odom])
• Moved get_param to __init__, to avoid creating zombie sockets with each call.
• Adding documentation to readme + image
• Added optional base frame to use to class init. Modified the return value of main clustering function to return pose message
• Migrated cluster_bounding_box_finder to hydro and set up test in node.
• Add package with node that tests recog message.
• Contributors: Bence Magyar, Sam Pfeiffer