EigenFace.py

# Necessary modules imported
import numpy as np
import sys
import cv2
import os

# Constants
dim = (32, 32) # cv2.imread(trainList[0], 0).shape

def loadListFromFile(filePath):
	return map(lambda path: path.strip(), open(filePath, 'r').readlines())

def train(trainList):
	trainset = [ np.ravel(cv2.imread(path, 0)) for path in trainList ]
	N = len(trainset)

	# Computing mean image
	mean = np.mean(trainset, 0)
	# print N, np.array(trainset).shape, mean.shape

	cov = lambda mat: np.dot(np.reshape(mat-mean, (-1, 1)), np.reshape(mat-mean, (1, -1)))
	C = (1.0/N)*np.array(np.sum((map(cov, trainset)), 0))

	w, v = np.linalg.eig(C)

	return mean, w, v

def computeCoeff(img, mean, eiVecs):
	imgR = np.ravel(cv2.resize(img, dim)) - mean
	coeff = map(lambda v: np.dot(v, imgR), eiVecs)
	return coeff

def computeLoss(tmpl, pred):
	# L2 Loss
	loss = float(np.real(np.sqrt(np.sum((pred-tmpl)*np.conj(pred-tmpl)))))
	return loss

# Incomplete
def test(testList, eiVecs):
	temp = map(lambda line: line.split(' '), testList)
	testset = [ np.ravel(cv2.resize(cv2.imread(path, 0), dim)) for path in temp[:, 0]]
	testLabels = temp[:, 1]

	output = [ computeCoeff(img, eiVecs) for img in testset ]
	data = load(tmplDir)

	# TODO : Compute loss and give a match
	return output

def saveTemplate(tmplDir, label, coeff):
	np.save(tmplDir + os.sep + label, coeff)

def loadTemplates(tmplDir):
	data = {}
	for name in os.listdir(tmplDir):
		if name.endswith(".npy"):
			data[name.replace(".npy", "")] = np.load(tmplDir + os.sep + name)

	return data

def loadModel(dataDir):
	return np.load(dataDir + os.sep + "mean.npy"), np.load(dataDir + os.sep + "model.npy")

def saveModel(dataDir, mean, eiVecs):
	np.save(dataDir + os.sep + "mean", mean)
	np.save(dataDir + os.sep + "model", eiVecs)