posedetection questions

classic Classic list List threaded Threaded
3 messages Options
Reply | Threaded
Open this post in threaded view
|

posedetection questions

Eohan
CONTENTS DELETED
The author has deleted this message.
Reply | Threaded
Open this post in threaded view
|

posedetection questions

Eohan
CONTENTS DELETED
The author has deleted this message.
Reply | Threaded
Open this post in threaded view
|

Re: posedetection questions

Rosen Diankov
Administrator
hi eohan,

When capturing the template, make sure you define the points clockwise
and the target template is big in the image.

Afterwards, try increasing the number of RANSAC iterations "--ransaciters=1000"

rosen,

2010/10/24 Eohan George <[hidden email]>:

> Hi,
> I am trying to use the posedetectiondb package to identify 6D poses of
> objects. I might be missing something here , however the following is what I
> do.
>
> Run camera1394 using the following command which publishes the
> images, rosrun camera1394 camera1394_node _guid:=814436100008622
> _camera_info_url:=file:///home/eohan/code/calibration/cameras/cameraUniBrain_640x480_rgb8.ini
>  _video_mode:=640x480_rgb8
> ROS_NAMESPACE=camera rosrun image_proc image_proc
> Run imagesift package, ROS_NAMESPACE=camera rosrun imagesift  imagesift
>
> At this point I have the following services and topics
> Service
> /camera/Feature0DDetect
> /camera/SetTemplate
> /camera/image_proc_1287859430985551755/get_loggers
> /camera/image_proc_1287859430985551755/set_logger_level
> /camera/imagesift/get_loggers
> /camera/imagesift/set_logger_level
> /camera/set_camera_info
> /camera1394_node/get_loggers
> /camera1394_node/set_logger_level
> /camera1394_node/set_parameters
> /rosout/get_loggers
> /rosout/set_logger_level
> Topic
> /camera/ImageFeature0D
> /camera/ObjectDetection
> /camera/camera_info
> /camera/image_color
> /camera/image_color/compressed
> /camera/image_color/theora
> /camera/image_mono
> /camera/image_mono/compressed
> /camera/image_mono/theora
> /camera/image_raw
> /camera/image_raw/compressed
> /camera/image_raw/theora
> /camera/image_rect
> /camera/image_rect/compressed
> /camera/image_rect/theora
> /camera/image_rect_color
> /camera/image_rect_color/compressed
> /camera/image_rect_color/theora
> /camera1394_node/parameter_descriptions
> /camera1394_node/parameter_updates
> /rosout
> /rosout_agg
> The problems are
> 1. However running the following to extract the features results in the
> error, though the pose extraction continues.
> $ ROS_NAMESPACE=camera rosrun posedetectiondb PointPoseExtraction.py
> [environment-core.h:966] ReadRobotXMLFile exception: openrave (Assert):
> [/usr/include/boost/smart_ptr/shared_ptr.hpp:418] -> T* boost::shared_ptr<
> <template-parameter-1-1> >::operator->() const [with T =
> OpenRAVE::InterfaceBase], expr: px != 0
> cannot create trimesh for :  openrave (Failed): [bool
> PyEnvironmentBase::AddKinBody(PyKinBodyPtr):2409]: invalid pointer
>
> 2.  The timings for detection are around 0.5-0.6 seconds on 4 GB machine
> with
> vendor_id : GenuineIntel
> cpu family : 6
> model : 23
> model name : Intel(R) Core(TM)2 Duo CPU     E8400  @ 3.00GHz
> cpu cores : 2
> cpu MHz : 3000.000
> cache size : 6144 KB
> Are there any hacks to speed up the detection?
> 3. The pose detection is quite off when you are closer to the object as
> shown in the images I've attached. I did try saving the features from
> different angles and distance without much help though. Any hacks to solve
> this ?
> http://img535.imageshack.us/i/openraveposedetectioner.png/
> http://img231.imageshack.us/i/openraveposedetectioner.png/
> 4. Are there any implementation currently available(in OpenRave or outside)
> to provide the CAD models for the object and allow detection with 6D pose of
> more intricate objects faster. (I know I'm asking for the ultimate answer
> for the best object detector :) , but people are creative, so I might have
> missed out stuff on google )
> I am attaching a screenshot of the rxgraph too.
> http://img191.imageshack.us/i/rxgraphposedetectioniss.png/
> Eohan
>
>
> ------------------------------------------------------------------------------
> Nokia and AT&T present the 2010 Calling All Innovators-North America contest
> Create new apps & games for the Nokia N8 for consumers in  U.S. and Canada
> $10 million total in prizes - $4M cash, 500 devices, nearly $6M in marketing
> Develop with Nokia Qt SDK, Web Runtime, or Java and Publish to Ovi Store
> http://p.sf.net/sfu/nokia-dev2dev
> _______________________________________________
> Openrave-users mailing list
> [hidden email]
> https://lists.sourceforge.net/lists/listinfo/openrave-users
>
>

------------------------------------------------------------------------------
Nokia and AT&T present the 2010 Calling All Innovators-North America contest
Create new apps & games for the Nokia N8 for consumers in  U.S. and Canada
$10 million total in prizes - $4M cash, 500 devices, nearly $6M in marketing
Develop with Nokia Qt SDK, Web Runtime, or Java and Publish to Ovi Store
http://p.sf.net/sfu/nokia-dev2dev
_______________________________________________
Openrave-users mailing list
[hidden email]
https://lists.sourceforge.net/lists/listinfo/openrave-users