Opencv_contrib: Error on training WBDetector

[thxmxx]

System information (version)

OpenCV => 3.1

Operating System / Platform => Ubuntu 14.04

Compiler => gcc

Detailed description

Simply trying to train a WaldBoost detector.
I have 722 png pos images 64x64 and 988 neg images also 64x64

This is my return:

compute features
pos=722 neg=988
i=   0 feat= 8179 thr=  0 casthr=2.693 alpha=2.693 err=0.005 loss=1.347e-01
neg 988/9
i=   1 feat=   36 thr=243 casthr=inf alpha=inf err=0.000 loss=inf
neg 9/0
OpenCV Error: Assertion failed (0 <= _rowRange.start && _rowRange.start <= _rowRange.end && _rowRange.end <= m.rows) in Mat, file /mnt/0ED4F1644821DB94/opencv/modules/core/src/matrix.cpp, line 477
terminate called after throwing an instance of 'cv::Exception'
  what():  /mnt/0ED4F1644821DB94/opencv/modules/core/src/matrix.cpp:477: error: (-215) 0 <= _rowRange.start && _rowRange.start <= _rowRange.end && _rowRange.end <= m.rows in function Mat

Aborted (core dumped)

Steps to reproduce

```.cpp Ptr<WBDetector> wbd = WBDetector::create(); wbd.operator->()->train("/home/thom/train/cars","/home/thom/train/non_cars"); ```

[berak]

you're running out of usable negatives

can you try to use larger negative images ? WBDetector tries to augment negative data by cropping multiple patches (in positive size) from the images.

[thxmxx]

Now it's failling on the first iteration...
3000(512x512) negative images

`
compute features
pos=1166 neg=11660
i= 0 feat= 381 thr= 0 casthr=inf alpha=nan err=-0.000 loss=nan
neg 11660/0
OpenCV Error: Assertion failed (0 <= _rowRange.start && _rowRange.start <= _rowRange.end && _rowRange.end <= m.rows) in Mat, file /media/thom/HD2/opencv/modules/core/src/matrix.cpp, line 477
terminate called after throwing an instance of 'cv::Exception'
what(): /media/thom/HD2/opencv/modules/core/src/matrix.cpp:477: error: (-215) 0 <= _rowRange.start && _rowRange.start <= _rowRange.end && _rowRange.end <= m.rows in function Mat

Aborted (core dumped)
`