trainAlgorithm.py

import cv2
import os
import fnmatch
import src.FileOperation as fo
import src.haar_cascade.haarcascade as hc
from matplotlib import pyplot as plt
import cmath

import config
import getCenter

filter_method_list=['mean','median','morph','none']
filter_size_list=[3,5,7,11,21]
edge_method_list=['canny','sobel','laplacian']
threshold1_list=range(10,60,5)
threshold2_list=range(30,150,10)
hough_dp_list=[1,1.5,2,2.5,3]

morph_kernel_rate_list=range(2,10,1)
threshold_method_list=['otsu','peak']
flag_precise=[True,False]

########################################################################
#Experiment Main
########################################################################
image_sets_name=[]
for root, dir, files in os.walk(config.DataSetPath+'.'):
    if root==config.DataSetPath+'.':
        image_sets_name=files

def ifInclude(circle,circle_truth):
    if circle[2]<circle_truth[2]:
        return 0
    dis=(circle[0]-circle_truth[0])**2+(circle[1]-circle_truth[1])**2
    boundary=(circle[2]-circle_truth[2])**2
    return 1*(dis<boundary)
def calculateDise(circle,circle_truth):
    return cmath.sqrt((circle[0]-circle_truth[0])**2+(circle[1]-circle_truth[1])**2)/circle_truth[2]

sum=0

for item in image_sets_name:
    print('image',item)
    truth=fo.readGroundTruth(item)
    src_img=cv2.imread(config.DataSetPath+item)
    count=0
    src=src_img.copy()
    (x,y)=getCenter.getCorneaCenter(src)
    center=(int(x),int(y))
    nRow=src_img.shape[0]
    tf=nRow/360
    cv2.circle(src,(int(x),int(y)),int(3*tf),(0,255,0),thickness=2*int(7*tf))
    sum+=calculateDise(center,truth)
    cv2.imwrite(config.resultSetPath+'ret_'+item,src)
    cv2.imshow('result',src)
    #cv2.waitKey(0)
    #sum+=calculateDise(point,truth)
    #print('loss',calculateDise(point,truth))
    print('')
result=sum/len(image_sets_name)
print(result)