Segfault when running Haartraining with OpenCV 2.2 (Bug #1174)

Added by Michael Spector over 13 years ago. Updated over 12 years ago.

Status:Cancelled Start date:
Priority:Low Due date:
Assignee:Marina Kolpakova % Done:

0%
Category:objdetect
Target version:2.4.3
Affected version: Operating System:
Difficulty: HW Platform:
Pull request:

Description

I'm getting segmentation fault when running a haartraining. Here's the debug information:

Backtrace:
-----------

(gdb) bt
#0 0x0000000000416ed5 in icvSortIndexedValArray_32s (array=0x7fffe004c880, total=246, aux=0x7ffff1612d60)
at /home/michael/opencv/opencv/modules/haartraining/cvboost.cpp:85
#1 0x000000000042df8d in cvCreateMTStumpClassifier (.omp_data_i=0x7fffffffaf80)
at /home/michael/opencv/opencv/modules/haartraining/cvboost.cpp:1034
#2 0x00007ffff5908fd2 in ?? () from /usr/lib/libgomp.so.1
#3 0x00007ffff5d1e971 in start_thread (arg=<value optimized out>) at pthread_create.c:304
#4 0x00007ffff4ec592d in clone () at ../sysdeps/unix/sysv/linux/x86_64/clone.S:112
#5 0x0000000000000000 in ?? ()

Values:
--------

(gdb) p *aux
$16 = {
data = 0x7fffe009d900 "\344\004Ҽ\225\254\232\275/\002\016\275vʇ\275\321\340\313;\255\255ἄ\361\365\274\317\020\065\035;\212\304\022\274\305\377\266\273q]\033<ɫ@\275\277\321\347<", step = 400}

(gdb) p array
$18 = (int *) 0x7fffe004c880

Please tell what information do you need from me.

Thanks!

History

#1
Updated by Vadim Pisarevsky over 13 years ago

it is recommended to turn off openmp when building haartranining.

#2
Updated by Michael Spector over 13 years ago

What method do you recommend for leveraging multi-process environment when running Haartraining, except for OpenMP?

#3
Updated by Vadim Pisarevsky over 13 years ago

haartraining is an old application, written and debugged on Windows using Intel compiler and its OpenMP in 2001-2003, I believe. That implementation worked fine on multi-core machines. Since then the application has been changed, some bugs have been fixed etc. So, I'm not sure that the latest version of haartraining will build and run fine even if it is built using Intel compiler.

Currently the only tested variant of haartraining with OpenMP enabled is haartraining from OpenCV 1.x built with Intel compiler.

We are planning to add TBB threading to our newer training application traincascade, but it will take some time (a few months, perhaps).

#4
Updated by Alexander Shishkov almost 13 years ago

  • Priority changed from High to Normal
  • Target version deleted ()

#5
Updated by Andrey Kamaev almost 13 years ago

  • Category changed from core to objdetect

#6
Updated by Alexander Shishkov almost 13 years ago

  • Priority changed from Normal to Low

#7
Updated by Alexander Shishkov almost 13 years ago

  • Assignee deleted (Vadim Pisarevsky)

#8
Updated by Alexander Shishkov almost 13 years ago

  • Target version deleted ()

#9
Updated by Marina Kolpakova over 12 years ago

haarcascade is too old application that is not longer supported. Instead of it we recommend to use traincascade application. It allows you to train Haad cascade as well as all other types of OpenCV cascades

  • Target version set to 2.4.3
  • Assignee set to Marina Kolpakova
  • Status changed from Open to Cancelled

Also available in: Atom PDF