Spym's profile - activity

Checked out this revision: https://github.com/ccny-ros-pkg/ccny_rgbd_tools/commit/88b1210524a32487f3538a5b7d28f2bbf3f6472e

Applied this patch: https://github.com/ccny-ros-pkg/ccny_rgbd_tools/issues/3

Compiled, everything is fine.

When I start nodes visual_odometry_node or keyframe_mapper_node I see the symbol lookup error appearing at run time:

[ INFO] [1361108695.958712716]: Starting RGBD Keyframe Mapper
/home/pavel/ros/ccny_rgbd_tools/ccny_rgbd/bin/visual_odometry_node: symbol lookup error: /home/pavel/ros/ccny_rgbd_tools/ccny_rgbd/lib/libccny_rgbd_registration.so: undefined symbol: _ZN9ccny_rgbd26removeInvalidDistributionsERKSt6vectorIN5Eigen6MatrixIfLi3ELi1ELi0ELi3ELi1EEESaIS3_EERKS0_INS2_IfLi3ELi3ELi0ELi3ELi3EEESaIS8_EERKS0_IbSaIbEERS5_RSA_
[visual_odometry_node-1] process has died [pid 30803, exit code 127, cmd /home/pavel/ros/ccny_rgbd_tools/ccny_rgbd/bin/visual_odometry_node __name:=visual_odometry_node __log:=/home/pavel/.ros/log/32aaa392-7908-11e2-8df2-00179a7bfadb/visual_odometry_node-1.log].
log file: /home/pavel/.ros/log/32aaa392-7908-11e2-8df2-00179a7bfadb/visual_odometry_node-1*.log

Ubuntu 12.04, GCC 4.6.3, ROS Fuerte.

Any help? Thanks in advance.

Hello.

Maybe this question is little offtopic here (sorry if so), I don't know where to ask it instead.

I need a method to detect the following kinds of objects using a stereo camera:

• any large obstacle, like building or tree, on a distance about 15m (tolerance doesn't matter, it's just for long-term collision avoidance)
• human body on distance about 7m
• a human's hand on distance about 2m

(Assuming good lighting, clear view, and so on)

Question for the people who have practiced with Bumblebee2: is this camera a good choice for these tasks? (http://www.ptgrey.com/products/bumblebee2/bumblebee2_stereo_camera.asp)

The Point Grey folks gave me nice guidelines about the stereo accuracy: http://www.ptgrey.com/support/kb/index.asp?a=4&q=103, but this isn't enough to make decision.

Thanks in advance,
Pavel.

UPD: Are there any other stereo cameras that are supported by ROS? A TOF camera seems to be a poor choice because of long ranges.

Hello.

I've installed http://playerstage.sourceforge.net/index.php?src=gazebo (Gazebo 0.10) from sources on Kubuntu 11.10 x64, and got a strange segfault on static objects initialization (before main()).

Used libraries: http://pastebin.com/4xAa1WcZ
GDB backtrace: http://pastebin.com/EiRY77Je
ltrace output: http://pastebin.com/9kMyAY67

The app fails here (gazebo-0.10.0/libgazebo/IfaceFactory.cc, static method):

using namespace gazebo;

std::map<std::string, IfaceFactoryFn> IfaceFactory::ifaces;

// Register a iface class.  Use by dynamically loaded modules
void IfaceFactory::RegisterIface(std::string classname, IfaceFactoryFn factoryfn)
{
  ifaces[classname] = factoryfn; // segfault in std::_Rb_tree_decrement(..)
}

Arguments are okay, stack is not overflowed.

Also look at the http://pastebin.com/0DXUNGgv (valgrind output): "Address 0x8 is not stack'd, malloc'd or (recently) free'd" - looks like invalid pointer, but arguments passed into the std::map are correct.

Used dependencies: bullet-2.79, fltk-1.1.10, ode-0.11.1, openal-soft-1.13, player-3.0.2, ogre-1.7.3;
flags for bullet: BUILD_SHARED_LIBS=ON, BT_USE_DOUBLE_PRECISION=ON.

What is my problem? Thanks in advance.

Here the results: http://pastebin.com/uAXXdSJq. Sadly, I see no corellation neither with frame switching nor number of keypoints. Maybe there are some other interesting parameters I may look closely?

It is easy to see that I shouldn't use anything but GFT. Also note that I used default parameters for each detector, except for max_range which I set to 15 meters. P.S. is there any mailing list or forum for the relevant discussions on ccny_rgbd?

Well, I did some benchmarking with pioneer_slam2 and pioneer_slam3: http://pastebin.com/28ewdVM4. I used native resolution 640x480, hence the poor performance of SURF detector. Images were published in real time, according to their timestamps.

Ivan, sorry for the offtopic, but I have a minor suggestion: it would be great to avoid direct stdout access from rgbd applications. Consider this: http://goo.gl/Kj39y, I suggest to replace printf() with ROS_DEBUG() here and everywhere else, to avoid the log pollution.

Yep, I already started following your repository on github, so I've seen an update on that issue. LSB stands for Least Significant Bit, i.e. amount of depth per one bit in our case.

About features: thanks, I'll give it a try a bit later today, stay tuned.

I suppose MONO16 is just alias for 16UC1. When I tried MONO16, I used it with resolution of 0.2mm per LSB, thereby VO discarded almost entire frame because depth values were above max_range threshold. I just changed the scale to 1mm/LSB and it that's it.

Thanks Miquel, but your wrapper fails in the same way, I've tested it. I made my own FOVIS wrapper because I needed the motion covariance, and had to do something with these NaNs as well.

Ivan, thanks it's working now, I converted the depth images into 16UC1 in millimeters. But anyway, in some parts of these datasets VO is unable to find enough features to track [http://pastebin.com/xFRg1V4c]. Maybe I need to switch to another kind of features?

Hi everyone,

Consider these pioneer datasets: http://vision.in.tum.de/data/datasets/rgbd-dataset/download#freiburg2_pioneer_slam2

They represent the specialties of my target application nicely, so I'm keen on testing some VO engines with them.

I've had a problem with CCNY RGBD: it works well with Kinect, but when I try to run it with Pioneer dataset it produces the following errors for each input frame:

[VO 147] Fr: 0.0 GFT[400][0]: 15.0 ICPProbModel[1]  0.1 TOTAL 15.2
[ WARN] [1361918845.707617958]: No features detected.
[ WARN] [1361918845.707702069]: Could not estimate motion from RGBD data, using Identity transform.
[VO 148] Fr: 0.0 GFT[400][0]: 13.9 ICPProbModel[1]  0.2 TOTAL 14.1
[ WARN] [1361918845.752454258]: [ICP] Not enough correspondences (0 of 15 minimum). Leacing ICP loop
[ WARN] [1361918845.752498383]: Could not estimate motion from RGBD data, using Identity transform.

I'm using the default settings. Color images are in BGR8 format, depth in 32FC1 or MONO16 (have tried both of them).

It is really strange that ccny_rgbd is unable to find any features on these images, I suppose there is some general issue, wrong image format maybe?