Difference between revisions of "FBM1 - “Object Perception”"

From RoCKIn Wiki
Jump to: navigation, search
(Undo revision 460 by BRSU (talk))
 
(5 intermediate revisions by one other user not shown)
Line 1: Line 1:
 
'''<pre style="color: red">To be updated for the 2015 RoCKIn Competition</pre>'''
 
'''<pre style="color: red">To be updated for the 2015 RoCKIn Competition</pre>'''
+
 
 +
== Location ==
 +
This benchmark takes place on the [[FBM table]]. The reference system for the table is defined by visual markers.
 +
 
 
== List of objects and associated frames ==
 
== List of objects and associated frames ==
moved to Section 5.1.2 (Feature Variation) + 5.1.3 (Associated Frames)
 
  
== Benchmark execution ==
+
The variation space for this FBM is represented by the set of objects that the robot may be presented with in the following table:
 +
 
 +
'''Object Classes/Instances''':
  
# An object of unknown class and unknown instance will be placed on a table in front of the robot
+
{| class="wikitable"
# The robot must determine the object’s class, its instance within that class as well as the 2D pose of the object w.r.t. the reference system specified on the table
+
|'''Object Classes'''
# The preceding steps are repeated until time runs out or 10 objects have been processed
+
|'''Object Instances'''
 +
|-
 +
!rowspan="4"| '''[a]''': Mugs
 +
|'''[a1]''': Large black mug [IKEA Code: 401.439.93]
 +
|-
 +
|'''[a2]''': Small white mug with yellow dots [IKEA Code: 702.348.97]
 +
|-
 +
|'''[a3]''': Coffee mugs [IKEA Code: 001.525.50]
 +
|-
 +
|'''[a4]''': Black jug [IKEA Code: 602.936.08]
 +
|-
 +
!rowspan="2"| '''[b]''': Forks and knives
 +
|'''[b1]''': Fork [IKEA Code 101.858.52]
 +
|-
 +
|'''[b2]''': Knife [IKEA code 901.929.62]
 +
|-
 +
!rowspan="2"| '''[c]''': Boxes
 +
|'''[c1]''': Yellow [IKEA Code: 200.474.50]
 +
|-
 +
|'''[c2]''': Pink [IKEA Code: 200.474.50]
 +
|-
 +
!rowspan="2"| '''[d]''': Picture frame
 +
|'''[d1]''': Gold-color frame [IKEA Code: 402.323.95]
 +
|-
 +
|'''[d2]''': Small black picture frame [IKEA Code: 601.674.93]
 +
|}
  
See the Rule Book for further details.
+
Variation space for object location is the surface of the benchmarking setup area where objects are located.
  
For each presented object, the robot must produce the result consisting of:
 
* object class name [string]
 
* object instance name [string]
 
* object localization (x [m], y [m], theta [rad])
 
  
Example of expected result:
+
'''Associated Frames''' (pictures with the corresponding representation can be downloaded [[Media:FBM1_frames.zip|"here"]]):
<pre>
+
object_class: a
+
object_name: a3
+
object_pose:
+
  x: 0.1
+
  y: 0.2
+
  theta: 1.23
+
</pre>
+
  
== List of variables to be logged ==
+
'''[a]''' Associated frame on the bottom plane of the mug with (see figures [[Media:FBM1_frames.zip|"here"]]) with:
 +
* Z-axis perpendicular to the “table” (surface upon which the object lies), pointed upward and coincident with the axis of symmetry of the mug (without taking into account the mug handle).
 +
* XY plane on the “table” with x-axis pointing to the mug handle.
  
The robot is required to log any sensor data used to perform the benchmark (e.g., images, point clouds).  The modalities for this are explained by [[Media:robotdata.txt|"this document"]].
+
'''[b]''' Associated frame on the bottom of the knife/fork (see figures [[Media:FBM1_frames.zip|"here"]]) with:
Only relevant data is expected to be logged (i.e., the pointcloud used to classify an object, more than one if an algorithm requiring multiple pointclouds is used).
+
* Z-axis perpendicular to the “table” (surface upon which the object lies), pointed upwards.
There are no restriction about the framerate: data can be saved, for the relevant parts of the benchmark, at the rate they are acquired or produced.
+
* XY plane on the “table” with x-axis pointing to the top part of the knife/fork.
The log may be a rosbag or the corresponding YAML representation, as specified in document: [[Media:Yaml_file_spec.pdf|"RoCKIn YAML Data File Specification"]].
+
  
The following are expected ROS topic names and corresponding data types:
+
'''[c]''' Associated frame on a top corner of the box (see figures [[Media:FBM1_frames.zip|"here"]]) with:
* object_class [<TT>std_msgs/String</TT>]: the recognized object class
+
* Z-axis perpendicular to the “table” (surface upon which the object lies), pointed upwards;
* object_instance [<TT>std_msgs/String</TT>]: the recognized object instance
+
* XY plane on the “table” with x-axis pointing to the smallest vertex of the box;
* object_pose2d [<TT>geometry_msgs/Pose2D</TT>]: the 2D pose of the recognized object
+
* X and Y axis must be coincident with vertices of the object (selection of the right corner);
* object_pose [<TT>geometry_msgs/Pose</TT>]: the 3D pose of the recognized object (if available)
+
* Note that the objects are not symmetrical. On one side there are a couple of hinges that teams should identify to remove ambiguities.
* image [<TT>sensor_msgs/Image</TT>]: sensorial data used to classify the object
+
* pointcloud [<TT>sensor_msgs/PointCloud2</TT>]: sensorial data used to classify the object
+
  
+Important!+ Calibration parameters for cameras must be saved. This must be done also for other sensors (e.g., Kinect) that require calibration, if a calibration procedure has been applied instead of using the default values (e.g., those provided by OpenNI).
+
'''[d]''' Associated frame on a corner of the picture frame plane (see figures [[Media:FBM1_frames.zip|"here"]]) with:
 +
* Z-axis perpendicular to the “table” (surface upon which the object lies), pointed upwards;
 +
* XY plane on the “table” with x-axis pointing to the smallest vertex of the picture plane;
 +
* X and Y-axis must be coincident with vertices of the object (selection of the right corner);
 +
* To remove the ambiguity, teams should identify the RoCKIn logo  (pictures for both frames can be downloaded [[Media:FBM1_frames.zip|"here"]]). (Commment from POLIMI: Should we specify it will be in a portrait manner?)

Latest revision as of 15:50, 16 November 2015

To be updated for the 2015 RoCKIn Competition

Location

This benchmark takes place on the FBM table. The reference system for the table is defined by visual markers.

List of objects and associated frames

The variation space for this FBM is represented by the set of objects that the robot may be presented with in the following table:

Object Classes/Instances:

Object Classes Object Instances
[a]: Mugs [a1]: Large black mug [IKEA Code: 401.439.93]
[a2]: Small white mug with yellow dots [IKEA Code: 702.348.97]
[a3]: Coffee mugs [IKEA Code: 001.525.50]
[a4]: Black jug [IKEA Code: 602.936.08]
[b]: Forks and knives [b1]: Fork [IKEA Code 101.858.52]
[b2]: Knife [IKEA code 901.929.62]
[c]: Boxes [c1]: Yellow [IKEA Code: 200.474.50]
[c2]: Pink [IKEA Code: 200.474.50]
[d]: Picture frame [d1]: Gold-color frame [IKEA Code: 402.323.95]
[d2]: Small black picture frame [IKEA Code: 601.674.93]

Variation space for object location is the surface of the benchmarking setup area where objects are located.


Associated Frames (pictures with the corresponding representation can be downloaded "here"):

[a] Associated frame on the bottom plane of the mug with (see figures "here") with:

  • Z-axis perpendicular to the “table” (surface upon which the object lies), pointed upward and coincident with the axis of symmetry of the mug (without taking into account the mug handle).
  • XY plane on the “table” with x-axis pointing to the mug handle.

[b] Associated frame on the bottom of the knife/fork (see figures "here") with:

  • Z-axis perpendicular to the “table” (surface upon which the object lies), pointed upwards.
  • XY plane on the “table” with x-axis pointing to the top part of the knife/fork.

[c] Associated frame on a top corner of the box (see figures "here") with:

  • Z-axis perpendicular to the “table” (surface upon which the object lies), pointed upwards;
  • XY plane on the “table” with x-axis pointing to the smallest vertex of the box;
  • X and Y axis must be coincident with vertices of the object (selection of the right corner);
  • Note that the objects are not symmetrical. On one side there are a couple of hinges that teams should identify to remove ambiguities.

[d] Associated frame on a corner of the picture frame plane (see figures "here") with:

  • Z-axis perpendicular to the “table” (surface upon which the object lies), pointed upwards;
  • XY plane on the “table” with x-axis pointing to the smallest vertex of the picture plane;
  • X and Y-axis must be coincident with vertices of the object (selection of the right corner);
  • To remove the ambiguity, teams should identify the RoCKIn logo (pictures for both frames can be downloaded "here"). (Commment from POLIMI: Should we specify it will be in a portrait manner?)