Main.py

# -*- coding: utf-8 -*-
"""

@author: Ankit Singh
"""

from ImageHandler import Imagehandler
from svmFile import svmClass
from utility import Pyramid,SlidingWindow
import yaml
import glob
import os
import random
import cv2 as cv

def App():
    with open("config.yml", 'r') as ymlfile:
        cfg  =  yaml.load(ymlfile)
    IOPlaces = cfg['Main']
    input = IOPlaces['Input']
    output = IOPlaces['Output']
#    print(output)
    directorypathpos = input['positive']
    os.chdir(directorypathpos)
    filesTypes = cfg['FileType']
    images = []
    for filetype in filesTypes:
        images.extend(glob.glob("*."+filetype))
    DataX = []
    DataY = [] 
    paths = [directorypathpos+image for image in images]
    
###adding and Computing HOG Vector
#   print(xrange(len(paths)))
    #image sizez are 96* 160
    
    for i in xrange(len(paths)):
#    for i in xrange(100):
#        Image = cv.imread(paths[i],cv.IMREAD_COLOR)
        obj = Imagehandler(paths[i])
        HogVector = obj.ImagesToTiles(16,16)
        DataX.append(HogVector)
        DataY.append(1)
        
    directorypathneg = input['Negative']
    os.chdir(directorypathneg)
    images = []
    for filetype in filesTypes:
        images.extend(glob.glob("*."+filetype))
    paths = [directorypathneg+image for image in images]
#   print(xrange(len(paths)))
    for i in xrange(len(paths)):
#    for i in xrange(10):
        Image = cv.imread(paths[i],cv.IMREAD_UNCHANGED)
        for j in xrange(10):
            rand = random.randint(0,50)
            img = Image[rand:rand+160,rand:rand+96]
            obj = Imagehandler(paths[i],img)
            HogVector = obj.ImagesToTiles(16,16)
#            print(len(HogVector))
            DataX.append(HogVector)
            DataY.append(0)
    
   
    
 ###train and test   

    svmObj = svmClass((DataX,DataY))
    svmObj.trainData()
    
    
####predict 
    os.chdir(output)
    images = []
    for filetype in filesTypes:
        images.extend(glob.glob("*."+filetype))
    paths = [output+image for image in images]
#    for i in xrange(len(paths)):
    for i in xrange(5):
        Image = cv.imread(paths[i],cv.IMREAD_UNCHANGED)
        imageHeight,imageWidth = Image.shape[:2]
        imageHeight = int(imageHeight/160)*160
        imageWidth = int(imageWidth/96)*96
        Image  =  cv.resize(Image,(imageWidth,imageHeight), interpolation  =  cv.INTER_CUBIC)
        for scaledImage in Pyramid(Image,2,(128,64)):
            for (x,y,window) in SlidingWindow(scaledImage,(14,14),(160,96)):
                if( window.shape[:2] !=  (160,96)):
                    continue
                oi = Imagehandler(paths[i],window)
                Hog = oi.ImagesToTiles(16,16)
#                print(len(Hog))
                val = svmObj.PredictData([Hog])
                val =  val[0]
                print(val)
                if val[1]> 0.8:
                    clone  =  scaledImage.copy()
                    cv.rectangle(clone, (x, y), (x + 160, y + 96), (0, 255, 0), 2)
                    cv.imshow("Window", clone)
                    cv.waitKey(1)
 

if __name__  ==  "__main__":
    
    cam = cv.VideoCapture(0)
    App()
    cam.release()
    cv.destroyAllWindows()