README.md

HaarCascade-Trained-to-detect-my-Watch-

I have trained to detect my watch using opencv and installation notes of opencv has been listed

In the program facecase.py is the main program

git clone https://github.com/arunodhayan/HaarCascade-Trained-to-detect-my-Watch-.git 

The code illustrates that

	1.The first few lines of code helps to download the images from imagenet.org
	2.Resize the image based on required pixel i choose(100X100).
3.Then there will be unwanted images to remove that create a folder called uglies and unwanted images will be deleted
	4.Now create a file bg.txt to list negative files for training 

To train the model

mkdir data 
mkdir info
1.opencv_createsamples -img typeimagename.jpg -bg bg.txt -info info/info.txt -pngoutput info -maxxangle 0.5 -maxyangle 0.5      -maxzangle 0.5 -num 1900 
    The numimages should be less than the negative image count(Eg if negative image is 1950 use 1900)
2.opencv_createsamples -info info/info.txt -num 1900 -vec positives.vec 
    To create positive vector files
3.opencv_traincascade -data data -vec positives.vec -bg bg.txt -numPos 1800 -numNeg 900 -numstages 12
    Here the positive must higher and negatives must be half