FullProgram.py

# -*- coding: utf-8 -*-
from ImageProcess import * 
from MachineLearning import *
import DensityAlignment
import math
    
def totalSVR(densityList, imageName,tileSize, overlap, trainingType): 
    """Does a complete run of the SVR learning algorithm. Takes in a training set of density
        data, an imageName in a string, the size of tile desired and an overlap as a percentage."""
        
    #trainingType: 0 = transect 
    #              1 = picList 
    #              2 = previous data set 
                  
    ##Note that if you have multiple transects you can simply save each transects group of pictures as a picList and then use method 1. 
    
                  
    if trainingType == 0: ###You want to pull data from a transect picture. 
        trainingData = numpy.zeros(50) ##Initialize the array to hold densities 
        
        ##Manually change densities where flowers existed here. All other entries will be 0. 
        ## This data comes from the BFS Survey Master Data Sheet. Values are # of flowers. 
        
        ###Transect 1 data set
    #    trainingData[0] = 5
    #    trainingData[2] = 14 
    #    trainingData[3] = 3
    #    trainingData[4] = 7
    #    trainingData[6] = 8
    #    trainingData[19] = 5
    #
    
    
        trainingData[23] = 253 
        trainingData[25] = 9
        trainingData[28] = 32
        trainingData[29] = 35 
        trainingData[30] = 4
        trainingData[34] = 515
        trainingData[35] = 14
        trainingData[36] = 299
        trainingData[39] = 1
        trainingData[40] = 51
        trainingData[43] = 179
        trainingData[45] = 10
        trainingData[47] = 2
        trainingData[48] = 3
        
        
        Start = (2839, 3449)
        End = (2496, 9760)
        
        #Start = (1035,588)
        #End = (456,1720)
        ## Change the name of the transect images, as well as the coordinates of the start and end in this function call. 
        
        imageList = DensityAlignment.divideTransect(Start, End,'TransectStitch2.jpg') ## Divide the transect into 50 images. Store in a list. 
        
        trainingData = list(trainingData)
        
        ###Transect 1 data
        #imageList = [ imageList[1], imageList[2], imageList[3], imageList[4], imageList[6], imageList[16], imageList[19]] + imageList[25:36]
        #trainingData= [ trainingData[1],trainingData[2], trainingData[3], trainingData[4], trainingData[6],trainingData[16], trainingData[19]] + trainingData[25:36]
   
        
        densityList = [int(i) for i in trainingData]
        ##Compute the metrics on each training image. 
        metricList, densityList = allTrainMetricsTransect(imageList, densityList)
        
        ### Save the training set  - metrics
        f = open('metricListTraining.txt', 'w')
        print(list(metricList), file=f)
        f.close()
        
        ### Save the training set - densities 
        f = open('densityListTraining.txt', 'w')
        print(densityList, file=f) 
        f.close()
        
    if trainingType == 1:  ##pull in pictures titled '1.jpg', etc. 
        NUMPICS = 100  ##Change the number of training pictures here. 
        imageList = makePicList(NUMPICS)
        
        imageList += ['TreeTest.jpeg'] 
        densityList = list(densityList)
        densityList += [0.0]
        

        
        
        metricList, densityList = allTrainMetrics(imageList, densityList)
        

        
        densityList = [int(i) for i in densityList]
        
        ### Save the training set  - metrics
        f = open('metricListTraining.txt', 'w')
        print(list(metricList), file=f)
        f.close()
        
        ### Save the training set - densities 
        f = open('densityListTraining.txt', 'w')
        print(densityList, file=f) 
        f.close()        
        
    if trainingType == 2:  #Just read in old data that was already saved.
        f = open('metricListTraining.txt', 'r')
        data = f.read()
        metricList = eval(data)
        
        g = open('densityListTraining.txt', 'r')
        data = g.read()
        densityList = eval(data)
        
    if trainingType == 3:  ### Other types of training sets. Manually enter here so that the otehr code doesn't need to change. 
        imageList = ['test1.jpg', 'test2.jpg']
        densityList = [1, 0]
        metricList , densityList = allTrainMetrics(imageList, densityList)
        
    ##Step 2: Scale the incoming training data 
    
    scaledTraining, scaler = scaleMetrics(metricList)  #Scale the training metrics to mean 0 and std 1 
    
    ### Train the machine learning algorithm 
    fit = svrAlg(scaledTraining, densityList)  #fit the algorithm 
    
    print('Machine Learning done')
    
    ###Calculate densities on full image 
    
    if True: #Make true if you need to calculate image metrics on a new image. Otherwise make false
        imageDens = allDensOverlap(tileSize, imageName, overlap, densityList, metricList, fit, scaler)
        fileName = imageName[0:-4] + 'Densities' +  '.txt'
        f = open(fileName, 'w')
        print(list(imageDens), file=f)
        f.close()
    else: 
        fileName = imageName[0:-4] + 'Densities' +  '.txt'
        f = open(fileName, 'r')
        data = f.read()
        imageDens = eval(data)

    print('Image densities computed')
    densMapShort(imageDens,imageName, overlap, tileSize)

def totalGauss(densityList, imageName,tileSize, overlap, trainingType): 
    """Does a complete run of the SVR learning algorithm. Takes in a training set of density
        data, an imageName in a string, the size of tile desired and an overlap as a percentage."""
    image = Image.open(imageName)
    imageSize = image.size
    
    #trainingType: 0 = transect 
    #              1 = picList 
    #              2 = previous data set 
                  
    ##Note that if you have multiple transects you can simply save each transects group of pictures as a picList and then use method 1. 
    
                  
    if trainingType == 0: ###You want to pull data from a transect picture. 
        print('Using single transect data for training')
        trainingData = numpy.zeros(50) ##Initialize the array to hold densities 
        
        ##Manually change densities where flowers existed here. All other entries will be 0. 
        ## This data comes from the BFS Survey Master Data Sheet. Values are # of flowers. 
        trainingData[2] = 14 
        trainingData[3] = 3
        trainingData[4] = 7
        trainingData[6] = 8
        trainingData[19] = 5
        
        ## Change the name of the transect images, as well as the coordinates of the start and end in this function call. 
        
        imageList = DensityAlignment.divideTransect((1035,588),(456,1720),'TransectStitch1.jpg') ## Divide the transect into 50 images. Store in a list. 
        densityList = list(trainingData)
        
        ##Compute the metrics on each training image. 
        metricList, densityList = allTrainMetricsTransect(imageList, densityList)
        
        
        ### Save the training set  - metrics
        f = open('metricListTraining.txt', 'w')
        print(list(metricList), file=f)
        f.close()
        
        ### Save the training set - densities 
        f = open('densityListTraining.txt', 'w')
        print(densityList, file=f) 
        f.close()
        
    if trainingType == 1:  ##pull in pictures titled '1.jpg', etc. 
        print('Using saved pictures for training data')
        NUMPICS = 50  ##Change the number of training pictures here. 
        imageList = makePicList(NUMPICS)
        metricList, densityList = allTrainMetrics(imageList, densityList)
        
        ### Save the training set  - metrics
        f = open('metricListTraining.txt', 'w')
        print(list(metricList), file=f)
        f.close()
        
        ### Save the training set - densities 
        f = open('densityListTraining.txt', 'w')
        print(densityList, file=f) 
        f.close()        
        
    if trainingType == 2: 
        print('Using previously calculated metric and density lists for training.')
        f = open('metricListTraining.txt', 'r')
        data = f.read()
        metricList = eval(data)
        
        
        g = open('densityListTraining.txt', 'r')
        data = g.read()
        densityList = eval(data)
        
    if trainingType == 3: 
        print('Using other type of training data')
        imageList = ['test1.jpg', 'test2.jpg']
        densityList = [100, 0]
        metricList , densityList = allTrainMetrics(imageList, densityList)
        
    ##Step 2: Scale the incoming training data 
    
    scaledTraining, scaler = scaleMetrics(metricList) 
    

    fit = gaussReg(scaledTraining, densityList)  ## Fit using Gaussian Regression
    
    
    print('Machine Learning done')
    
    ###Calculate densities on full image 
    
    if True: #Make true if you need to calculate image metrics on a new image. Otherwise make false
        imageDens = allDensOverlap(tileSize, imageName, overlap, densityList, metricList, fit, scaler)
        fileName = imageName[0:-4] + 'Densities' +  '.txt'
        f = open(fileName, 'w')
        print(list(imageDens), file=f)
        f.close()
    else: 
        fileName = imageName[0:-4] + 'Densities' +  '.txt'
        f = open(fileName, 'r')
        data = f.read()
        imageDens = eval(data)

    print('Image densities computed')
    densMapShort(imageDens,imageName, overlap, tileSize)
 
    
def makePicList(numSites): 
    """makes an image name list for a given number of sites.""" 
    nameList = []
    for i in range(numSites): #for each site 
        currentName = str(i+1)+".jpg" 
        nameList += [currentName]
    return nameList
        
#if __name__ == "__main__":
    
    #totalSVR(['SuperSmallTile.jpg'], [0.5], 'SmallTile.jpg', 75, 0.5)