deer detection 2
Haartraining
1. gather positive images
I gathered over 50 deer images, and duplicated and retouched to silhouette. I would use infra-red camera, silhouette might be good. So I got 250 deer images.
Through image cropper program, make to “info .txt” that contains cropped positions of all positive images. This stage is so tired, I had to drag to make deer position 250 times.
2. createsamples.exe
this step is converting info .txt to *.vec file. vec will need for last step.
3. gather negative images
I gather over 300 negative images. Negative image means any image doesn’t include deer. It might be need to compare with positive images. list on “infofile.txt”
4. haartraining.exe
This step takes lots of time, not just few hours. I used only 600 images, so It took half days, but some people do with 10,000 images, taking one week. This step makes bunch of xml files.
5. convert.bat
The last step, it combines all xml files to one haarcascade xml file.
Done!
