Informaticus

Today I fixed a memory bug in one of my programs. Albeit I always release memory cvReleaseImage after allocating it with cvCreateImage, memory consumption was enormous. Commenting out some of the source code quickly showed the responsible code fragments. Using cvCreateImage and afterwards cvCloneImage made it impossible to release the initially allocated memory. Therefore after some iterations (depending on your main memory) this program will crash. To fix this issue simply replace the cvCloneImage with cvCopy.

Bad code:

DO
 img2 = cvCreateImage(cvSize(img->width, img->height), img->depth, img->nChannels);
[...]
 img = cvCloneImage(img2);
 cvReleaseImage(&img2);
LOOP

Good code:

DO
 img2 = cvCreateImage(cvSize(img->width, img->height), img->depth, img->nChannels);
[...]
 //img = cvCloneImage(img2);
 cvCopy(img2,img,NULL);
 cvReleaseImage(&img2);
LOOP

I want to share my collection of links of how to build a classifier based on haar-like feature using OpenCV. If you know another (open source) implementation please leave me a message!

An example where a bowl is detected
http://lab.cntl.kyutech.ac.jp/~kobalab/nishida/opencv/OpenCV_ObjectDetection_HowTo.pdf

A wiki describing how to build a haar classifier for robotic vision
http://cgi.cse.unsw.edu.au/~cs4411/wiki/index.php?title=OpenCV_Guide#Haar_Classifier

Frequently asked questions about haar training (mainly basics about positive and negative input file format, how to read the output of the OpenCV haar training program ...)
http://www.computer-vision-software.com/blog/2009/11/faq-opencv-haartraining/

Short summary of the output during haar training
http://cv-kolaric.blogspot.com/2008/03/output-from-haartrainingexe.html

How to fix cvboost.cpp if you encounter a "bad flag" error during haar training
http://tech.groups.yahoo.com/group/OpenCV/message/69620

Another example of how to detect a bowl
http://www.computer-vision-software.com/blog/2009/06/opencv-haartraining-detect-objects-using-haar-like-features/

One of the best tutorial (imho) of how to build a cascade of boosted classifiers based on haar-like features. It's a step by step manual for face detection.
http://note.sonots.com/SciSoftware/haartraining.html

I spent the last days restarting, updating and googling to get my IPod Nano (5G) working on Windows XP (+ SP3). It was not recognized by Windows nor by iTunes. To get it work properly I had to

• start "msconfig",
• choose the "Startup" tab,
• "Selective Startup",
• "Service" tab,
• check "Hide Microsoft services",
• disable all services except "Apple Mobile Device" and "IPod Service",
• and finally reboot windows.

Finally Windows XP found the IPod, reformated it with vfat32 and from that moment on it was correctly recognized everytime it was connected to the computer.

My Easy Peasy Linux was lost after installing Windows 7. Windows couldn't setup (by default) a boot manager so I had to recover my boot record. My Linux knowledge is quite good I thought but after a while I recognized that I missed a development in the last years - udev.  I used a mini linux-distribution stored on my USB stick (my EeePC doesn't have a CD drive) to boot into Linux, mounted the installed Easy Peasy and restored the boot record using grub. Now I have a handy menu where I can choose between Windows 7 and Easy Peasy.

Steps to recover a lost Linux system after installing Windows:

Assuming that Linux is installed on /dev/sda1, Windows on /dev/sda3.

sudo mkdir /mnt/root
sudo mount  /dev/sda1 /mnt/root
sudo mount -t proc none /mnt/root/proc
sudo mount -o bind /dev /mnt/root/dev
sudo chroot /mnt/root /bin/bash
edit your menu.lst[1]
sudo grub-install /dev/sda[2]

Lines 3 and 4 are essential, without them Grub cannot be installed on a system using udev.

[1]
In case of Windows you have to add something like this:

title win7
rootnoverify(0,2)
chainloader +1

Please check the menu.lst somewhere in the upper part of the file is an example configuration for Windows! Further (0,1) might not be the correct partition, replace it with your partition where Windows has been installed.

[2]
Check your installation!