LINE-MOD crashes maybe due to libosmesa6? #18
Comments
theNerd247
changed the title
|
Hello, I'm using Ubuntu 14.04.3 and it works well for me. |
|
Yes, here's what I ran: |
|
ok, |
|
Yes they're at the newest version also. Are you using the ros-indigo-object-recognition-linemod package or are you building from the latest git source? |
|
yeah I'm building from source all of these packages, but I will check the release and let you know |
|
I'm using a kinect2 with the iai_kinect2 ros package. |
|
Any luck? |
|
Have you solved the problem?? I am facing the exact same problem now |
|
@theNerd247 Hi, can you try compiling ork_renderer (a dependency of LINEMOD) with GLUT instead of osmesa? Here is a discussion in ORK maillinglist about the same issue. |
|
Sadly I can't currently. The project that used linemod was with work and I won't be able to work on the same system till this summer. |
|
Hi all, |
|
Can any of you please detail the different instructions you are launching ? For ORK or for the camera. Thx. |
|
Because I'm using ROS, so the instructions are the ROS part on the ORK tutorial, and especially the linemod which got me the error: http://wg-perception.github.io/ork_tutorials/tutorial03/tutorial.html The specific instruction I used which caused the problem is the detection part in linemod tutorial above, i.e., rosrun object_recognition_core detection -c I met no problem before detection, with the given coke model and a coke. Thanks! |
|
I am using exactly the same steps for running LINEMOD. On January 1, 2016, at 6:03 PM, YantianZha [email protected] wrote: Because I'm using ROS, so the instructions are the ROS part on the ORK tutorial, and especially the linemod which got me the error: http://wg-perception.github.io/ork_tutorials/tutorial03/tutorial.html The specific instruction I used which caused the problem is the detection part in linemod tutorial above, i.e., rosrun object_recognition_core detection -c rospack find object_recognition_linemod/conf/detection.ros.ork. I met no problem before detection, with the given coke model and a coke. Thanks! — |
|
And I'm sure my libfreenect driver ran ok. Because I got the right value from the four subscribed camera topics (you would see them after rosrun object_recognition_core detection -c rospack find object_recognition_linemod/conf/detection.ros.ork) |
|
ell, I cannot reproduce with Jade on Vivid. Can you please trace down the error by using gdb ? You can do so with ecto like that: |
|
The tutorial requires me to add the namespace of cell, but the cell name below is "pipeline1", which is actually from the conf file of detection (https://github.com/wg-perception/linemod/blob/master/conf/detection.ros.ork)? I'm new to these ecto/ork things. So can you give me more suggestions? Thanks! Starting program: /usr/bin/python /home/yantian/ws/src/ork_core/apps/detection -c /home/yantian/ws/src/ork_linemod_new/conf/detection.ros.ork [function_name] = process_with_only_these_inputs [type] = std::bad_alloc [what] = std::bad_alloc [when] = While triggering param change callbacks [Thread 0x7fffd1a75700 (LWP 16375) exited] |
|
@YantianZha Do you think you can provide us a rosbag of your camera's capture (point cloud and rgb and tf) so that we could debug LINEMOD with your setting? Thanks. |
|
Thanks @hris2003 . Here is the link to my bag file: https://drive.google.com/file/d/0B6lGtF9IuZWCV2tzTHl6bXRFemM/view?usp=sharing Then after training the coke model, using "rosrun object_recognition_core detection -c |
|
Hi, So I run ork_linemod with your bag, and it works. No errors, no warning. For your problem, I suspect either something's wrong with the database or
Good luck. Ha Dang. On Sat, Jan 2, 2016 at 6:22 PM, YantianZha [email protected] wrote:
|
|
Some lines in https://groups.google.com/d/msgid/object-recognition-kitchen/12546f19-4033-4ecf-a854-54d91bebf73d%40googlegroups.com need to be changed: Then go to src/CMakeLists.txt and block these few line in the header #if USE_GLUT "Then go to src/CMakeLists.txt and block these few line in the header" --> " |
|
For 6, I meant run the freenect.launch. And yes, the concerned file is renderer3d.cpp ([email protected] did send a Ha Dang. On Sat, Jan 2, 2016 at 10:10 PM, YantianZha [email protected]
|
|
I just did everything again, including reinstalling ecto. The error still emerged. Different from my previous trial using Further, I met same error (ecto.cellException) in using "rosrun object_recognition_capture orb_template -o my_textured_plane", following http://wg-perception.github.io/object_recognition_core/quickguide.html#detect-objects. But considering that the linemod tutorial did not mention using "object_recognition_capture", I don't know whether this matters. |
|
I think it matters as it depends on by which library package you are creating your database in the training phase. (osmesa or glut). If you did not created your database with glut, then linemod would be generating the same error message you received. The linemod is not working currently with osmesa. Please refer to this page where I have tried to summarize how we can run the linemod. Thank you. |
|
@YantianZha Do you think you can try ork on a different computer where there's not yet ork installed? |
|
Yes it worked. Thank you so much all guys! And @jontromanab , I'm sure I deleted the object firstly as what you suggested in your post. But I did not delete the overall data base (the mesh is still in it, delete object cmd can't delete the mesh) I think the reason is, while I changed a list of files in renderer folder, and then ran catkin_make, my system still "make with the previous configuration". For example, you may note before that, if you just change a little part in a "certain place" in a catkin package, catkin_make would not actually "see" it, but just passed the make process. Thanks again :) |
|
Hi @jontromanab, |
|
Hello @YantianZha ....Sorry for the late reply. Quite busy with the other stuffs. Yes. I think I have trained one coke can and one pringles chip box and it detected them separately. |
|
Hi @YantianZha Have you got any solution yet? |
|
try OSMesa catkin_make --cmake-args "-DUSE_GLUT=OFF" |
|
I have the exact problem described in this thread, I can't however run the training after compiling from source training gives error: [cell_type] = ecto::py::BlackBox [function_name] = process_with_only_these_inputs [type] = std::runtime_error [what] = exception_type CellException [cell_type] = ecto_linemod::Trainer [function_name] = process_with_only_these_inputs [type] = std::runtime_error [what] = value type is 172215184 not 2 Sounds like there's a loop running wild. The bold value changes every time I try to run the program. Any ideas on what might be the problem? |
|
@theNerd247 , @jontromanab Did you guys ever succeeded in using ORK with the Kinectv2 and libfreenect2 ? |
|
@trykks Have you been able to resolve the issue? I'm having the same problem! |
|
I'm afraid not. I abandoned it completely due to lack of time in our project. |
|
This is due to mesa issues that hasn't been solved yet. |
|
@JimmyDaSilva I've done what you said and it doesn't work! do you have any other suggestion? |
|
Well what is the error exactly? |
|
the error is this one: Training 2 objects. [cell_type] = ecto::py::BlackBox [function_name] = process_with_only_these_inputs [type] = std::runtime_error [what] = exception_type CellException [cell_type] = ecto_linemod::Trainer [function_name] = process_with_only_these_inputs [type] = std::runtime_error [what] = value type is 50032384 not 2 that's when I'm trying to train the objects. 2016-05-04 10:50 GMT+01:00 Jimmy DA SILVA [email protected]:
|
|
I am no expert here since the error doesn't help a lot but I can try to help. Can you please:
And tell me which training fails. |
|
Now it's working. I would be able to train the objects. I install it from [ INFO] [1462362900.067559485]: Initialized ROS. node_name: do you know what's happening here? thanks in advance 2016-05-04 12:43 GMT+01:00 Jimmy DA SILVA [email protected]:
|
|
by the way, the training from the source it's is not working, I'm having 2016-05-04 12:58 GMT+01:00 Marti de castro cros <
|
|
I'm having the same error everytime I try to train. Training 1 objects. [cell_type] = ecto::py::BlackBox [function_name] = process_with_only_these_inputs [type] = std::runtime_error [what] = exception_type CellException [cell_type] = ecto_linemod::Trainer [function_name] = process_with_only_these_inputs [type] = std::bad_alloc [what] = std::bad_alloc any other ideas? thanks 2016-05-04 13:00 GMT+01:00 Marti de castro cros <
|
|
So are you using ORK from the sources or from the release ? |
|
I've try in both ways. From release it works properly. Then, I have try to 2016-05-04 15:28 GMT+01:00 Jimmy DA SILVA [email protected]:
|
|
Alright keep using the sources then. |
|
yes, I'm sure about that. 2016-05-04 15:52 GMT+01:00 Jimmy DA SILVA [email protected]:
|
|
To find all the objects still existing: That way we will make sure everything has been deleted (mesh, trained data, etc...) |
|
I met the same problem,line 24, in run_plasm(args, plasm) |
|
@theNerd247 @nlyubova @hris2003 @YantianZha [cell_name] = pipeline1 [function_name] = process_with_only_these_inputs [type] = std::bad_alloc [what] = std::bad_alloc [when] = While triggering param change callbacks |
Currently LINE-MOD is crashing with the following error:
I ran the current master branch on another machine (Ubuntu 14.04.2) which hasn't been updated since August 2015 and linemod works fine. So then I grabbed all the packages that updated on my system and cross referenced it with the list of system dependencies that
rosdepshows forobject_recognition_linemodand libosmesa6-dev is the only package that is showing up. So I'm assuming a system update to libosmesa6-dev is causing the problem. However downgrading to libosmesa6-dev.10.1.0-4ubuntu5 doesn't seem to help anything.The text was updated successfully, but these errors were encountered: