ObjectRecognizer hangs after training?

[chyphen777]

I tried to run ObjectRecgnizer, first time, using the example command from the tutorial, and the process seemed to hang (for over an hour) during or after the training. sift and views associated with each of all TUW models have been generated. I tried it using code from both the latest version from https://rgit.acin.tuwien.ac.at/v4r/v4r/ and an earlier version from https://github.com/strands-project/v4r, and both of them hanged after training. I am wondering if either of them is waiting for input.

Computing boundary points took 2ms.
Training sift (with id "") on view /yellow_toy_car/data/TUW/TUW_models/yellow_toy_car/views/cloud_00000019.pcd took 124ms.
Plane extraction took 31ms.
Feature Encoding took 120ms.
Computing boundary points took 8ms.

---

Solution: 11
Final cost: -7.48687
Number of evaluations: 5
Number of accepted moves: 3

---

I did kill it and reran the command, and it successfully loaded trained data, but it still hanged.

cchien@cchien:~/rgbd/v4r$ ./build/bin/ObjectRecognizer -m data/TUW/TUW_models -t data/TUW/test_set -z 2.5 --or_do_sift true --or_do_shot false --or_remove_planes 1 -v
E1003 16:55:03.733016 21942 ObjectRecognizerParameter.h:150] The given parameter for feature types, classification methods and configuration files for global recognition are not the same size!
E1003 16:55:03.733120 21942 ObjectRecognizerParameter.h:155] Global recognition is disabled!
Plane extraction took 31ms.
Feature Encoding took 405ms.
Computing boundary points took 7ms.

---

Solution: 11
Final cost: -7.48687
Number of evaluations: 5
Number of accepted moves: 3

---

I did notice a message "Global recognition is disabled". Not sure if it is related. And I missing something. Any suggestions for making ObjectRecognizer working will be appreciated.

Thanks, CC

[ToMadoRe]

From your output it seems that it already finished recognition and accepted two object hypotheses. As you turned on visualization (-v), it should open a PCL visualization window where it shows the detected objects. You can press 'r' on that window to reset the viewpoint and 'e' (or just close the window) to exit. The recognizer will wait until you close the window to process the next frame. If you want to turn off visualization, just run the command without -v.

[chyphen777]

Thanks.  That was what I was wondering, but not sure.  I couldn't find output (pose) files even though I had -o output.txt at the command line.  I will try to figure it out.  Thanks for clarification.  CC. On Wednesday, October 4, 2017 3:34 AM, Thomas Fäulhammer <[email protected]> wrote: From your output it seems that it already finished recognition and accepted two object hypotheses. As you turned on visualization (-v), it should open a PCL visualization window where it shows the detected objects. You can press 'r' on that window to reset the viewpoint and 'e' (or just close the window) to exit. The recognizer will wait until you close the window to process the next frame. If you want to turn off visualization, just run the command without -v.— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub, or mute the thread.