 |
training images
nandhini from India [1 posts] |
12 year
|
hi,
i tried to train a simple image from a folder but roborealm closes when i start training... what is the solution for this?
|
|
 |
jpiz from Mexico [3 posts] |
12 year
|
I have the same problem, some format exists in specific of the images, always on having proved the program it generates a mistake and is closed, I could not have loaded the image to prove this application. Any help in this topic would be grateful for them to him.
Regards
|
|
|
|
Anonymous |
12 year
|
Nandhini,
Can you post the image here that you are trying to train on? Most likely there is an issue with that particular image.
STevn.
|
|
|
|
jpiz from Mexico [3 posts] |
12 year
|
I have 12 objets that all whit a "S" like the image...
|
|
|
|
Anonymous |
12 year
|
Thanks, go ahead and download 2.42.18 and give that version a try.
Let me know if you continue to see any issues.
Keep in mind that the module expects a binary (black and white) image which you can create via thresholding the image.
STeven.
|
|
|
|
jpiz from Mexico [3 posts] |
12 year
|
Thanks Steven, now the problem has disappeared ... the image is loaded without problem, and the program is not closed.
thank you very much
Juan Pablo
Regards...
|
|
|
|
Anonymous |
12 year
|
hi steve,
i tried with the image 's' in previous post,still the problem persists.
i am using v2.42.21.
|
|
|
|
Anonymous |
12 year
|
|
can i get the steps you follow to do this process? just to ensure that i am working correctly.
|
|
|
|
Anonymous |
12 year
|
Anonymous,
Do you see a c:\RoboRealm.log file after the crash is generated? Can you zip and post that here?
The process we are using is very basic. Add Shape_Match to empty pipeline. Select Train button with single image above in it and train.
STeven.
|
|
|
|
Anonymous |
12 year
|
i too followed the same steps.
i dint see any such files after crashing.
mine is a trial version is this a reason that this module is not working? i will own this software as soon as possible!
Nandhini
|
|
|
|
Anonymous |
12 year
|
Nandhini,
Can you verify what version you are running? The error was only recently fixed (along with the new error reporting that generates c:\RoboRealm.log) so it appears you need to update your copy. Use the original link you received via email or click on the Download button in the Options dialog.
STeven.
|
|
|
|
Anonymous |
12 year
|
hi steve,
i am using v2.42.21 . i dint get any error report when the software closes .
i am using the original link sent by RoboRealm.
|
|
|
|
Anonymous |
12 year
|
Hmm, perhaps something else is going on that we don't understand. Can you instead follow the steps on #27 at
http://www.roborealm.com/FAQ.php#q21
and send us that debug file (its also RoboRealm.log but is generated while running rather than on a crash).
Thanks,
STeven.
|
|
|
|
Anonymous |
12 year
|
hi steve,
i have attatched the .log file while working in shape match module.
my object recognition module is working properly. so i started working on that.
i need another clarification too. Is it possible to give X,Y,Z orientation of an object from Roborealm to input of Forward and Reverse Kinematics?
 RoboRealm.zip
|
|
|
|
Anonymous |
12 year
|
Nandhini,
Thanks for the log file. That answered two of our questions:
1. The version you are using .16 was prior to the crash debug file creation in .17 That's the reason you did not see that file.
2. The above comments are about the shape_match module and NOT the image_match module which you are using. Its naturally important to know which module you are actually having issues with.
We've corrected the detected problems with that module and have uploaded a new version. Please verify that the title bar reads 2.42.25 after downloading the latest copy before you begin your tests.
Note that the interface to that module has been altered so it will look a bit different.
Thanks,
STeven.
|
|
|
|
Anonymous |
12 year
|
In reference to your last question, yes in a calibrated system (i.e. camera sees the same worksapce as robot) the XY and Z (orientation not depth) can be given to the robot. If the camera and robot do not share the same workspace (i.e. 0,0 in the image does not align to 0,0 in the robot space) then you will have to convert from one to the other before sending the robot its coordinates.
STeven.
|
|
|
|
Anonymous |
12 year
|
|
|
|
|
Anonymous |
12 year
|
hi
our robot is a humanoid which we are developing for restaurant serving purpose.
eg: if the robo finds a plate, it has to go and pick it up. here it is necessary that the robo must recognize the object as a plate in which ever orientation it lies in.
so i need to train the robo about the plate in all possible orientations.
also i need to teach the socket image for self charging purpose.
is it possible to do in RR?
sorry for posting shape match .log file. since both shape and image matching were not working properly ,i posted one of it. now i updated the software.
Thanks
Nandhini
|
|
|
|
Anonymous |
12 year
|
Nandhini,
Detecting plates at any orientation esp on a white tablecloth is a very difficult problem and not something that we have a module in RR to just do.
If you can use a different color tablecloth than the plates that will be easier to detect as you would then be looking for oval shapes (assuming a non-top down view of the table) that can serve as the initial extraction for plates.
Have a look at the attached for an example of an ideal view.
STeven.
 program.robo
|
|
This Thread