Introducing
Your new presentation assistant.
Refine, enhance, and tailor your content, source relevant images, and edit visuals quicker than ever before.
Trending searches
superquadric-model
8
Theory
Researcher of the Week:
superquadric-model module
Superquadric fuctions
inside-outside function:
- F > 1: point outside
- F < 1: point inside
- F = 1: point on surface
5 parameters for shape
6 parameters
for pose
Optimization problem
Goal
Noteworthy ...
Future works
coding & applications
- shape indipendence
- minimum volume
Superquadric visualization
You can find
all the information about superquadric-model
module on the github repo:
Noisy point clouds
Object detection and modeling
Wrong solution
(at first glance)
iCub looks at the object
VTK/OpenGL
the object is detected
Thank you for your attention!
a 3D model is computed
the model is visualized
Grasping application
What the robot sees..
Possible solutions
..any questions or comments?
..what actually the robot sees
https://github.com/robotology/superquadric-model
Same cost function value!
Partial non-noisy clouds:
Requests:
Pipeline
Detection
Good grasping pose
Modeling
Visualization
Ideas
Online Pipeline
... and everything robust and in real-time
Compute
superquadric
from 2D
to 3D
blob
extractor
Seed point
Merged point clouds
(ICP)
What about
one-shot behaviour?
LbpExtract output
3D points sampled
on the superquadric
Partial point cloud
From 3D points to 2D Pixels
2D blob of the
selected object
SFM
3D point cloud
(Voxelization)
Point cloud time refinement and filtering
which is
box id?
superquadric-module
it's 0!
show
superquadric
OPC
ok, now give me
the 2D seed point
of object id 0
superquadric-module
Filtering (optional)
Here it is!
OPC
Object name
Trajectory planning
Code
2D surface overlapping
Different optimization problem formulation
Offline Pipeline
Why filtering?
Before ...
Point cloud file
Filtering or not?
... after!
superquadric
computation
Dependencies
Code
... What happens next?
Then, let's try our code!
Obstacle avoidance
Dependencies:
- YARP
- iCub
- OpenCV
- IPOPT
Interior Point OPTimizer
External application
(Interactive Object Learning)
- IOL:
Connections
(Local Binary Pattern Extraction)
(Structure from Motion)
- lbpExtract
- SFM
- OPC
(Object Property Collector)
Code
How to combine everything?
superquadric-
model
IOL
yarpviewers
cameras
left camera
right camera
superquadric-model
rpc
yarpview
lbpExtract
rpc
SFM
OPC
RPC Thrift services
Set and/or get info about:
- object name/seed point
- visualization color
- exploited eye
- maximum number of points for superquadric calculation
- number of visualized points
- superquadric parameters
- plot options (points or superquadric)
- advanced options (filtering & IPOPT)
Code
Giulia Vezzani - 13/05/2016