def mvndSkinDetection():
'''
Skin detection - compute a MVND for both skin and non-skin.
Classify the test and training image using the classify helper function.
Note that the "mask" binary images are used as the ground truth.
'''
sdata = scipy.io.loadmat(os.path.join(dataPath,'skin.mat'))['sdata']
ndata = scipy.io.loadmat(os.path.join(dataPath,'nonskin.mat'))['ndata']
mvn_sskin = MVND(sdata)
mvn_nskin = MVND(ndata)
# Optain priors
mask = imageHelper()
mask.loadImageFromFile(os.path.join(dataPath, 'mask.png'))
prior_skin = mask.getLinearImage().mean()
prior_nonskin = 1-prior_skin
print("TRAINING DATA")
trainingmaskObj = imageHelper()
trainingmaskObj.loadImageFromFile(os.path.join(dataPath, 'mask.png'))
trainingimageObj = imageHelper()
trainingimageObj.loadImageFromFile(os.path.join(dataPath, 'image.png'))
classify(trainingimageObj, trainingmaskObj, mvn_sskin, mvn_nskin, "Training", prior_skin, prior_nonskin)
print("TEST DATA")
testmaskObj = imageHelper()
testmaskObj.loadImageFromFile(os.path.join(dataPath, 'mask-test.png'))
testimageObj = imageHelper()
testimageObj.loadImageFromFile(os.path.join(dataPath, 'test.png'))
classify(testimageObj, testmaskObj, mvn_sskin, mvn_nskin, "Test", prior_skin, prior_nonskin)
plt.show()
def gmmSkinDetection() -> None:
'''
Skin detection - train a GMM for both skin and non-skin.
Classify the test and training image using the classify helper function.
Note that the "mask" binary images are used as the ground truth.
'''
K = 1
iter = 50
sdata = scipy.io.loadmat(os.path.join(dataPath, 'skin.mat'))['sdata']
ndata = scipy.io.loadmat(os.path.join(dataPath, 'nonskin.mat'))['ndata']
gmms = gmm_em(sdata, K, iter)
gmmn = gmm_em(ndata, K, iter)
print("TRAINING DATA")
trainingmaskObj = imageHelper()
trainingmaskObj.loadImageFromFile(os.path.join(dataPath, 'mask.png'))
trainingimageObj = imageHelper()
trainingimageObj.loadImageFromFile(os.path.join(dataPath, 'image.png'))
prior_skin, prior_nonskin = get_prior(trainingmaskObj)
classify(trainingimageObj, trainingmaskObj, gmms, gmmn, "training", prior_skin=prior_skin,
prior_nonskin=prior_nonskin)
print("TEST DATA")
testmaskObj = imageHelper()
testmaskObj.loadImageFromFile(os.path.join(dataPath, 'mask-test.png'))
testimageObj = imageHelper()
testimageObj.loadImageFromFile(os.path.join(dataPath, 'test.png'))
classify(testimageObj, testmaskObj, gmms, gmmn, "test", prior_skin=prior_skin, prior_nonskin=prior_nonskin)
plt.show()
def classify_data(self):
# helper function to call the mthods in the Cluster Class
trainingFileName = "training.csv"
testingFileName = "testing.csv"
outPutFileName = "prediction.csv"
cla = classify(trainingFileName, testingFileName, outPutFileName)
cla.NaiveBayesClassifier()
return outPutFileName, trainingFileName
def mvndSkinDetection() -> None:
'''
Skin detection - compute a MVND for both skin and non-skin.
Classify the test and training image using the classify helper function.
Note that the "mask" binary images are used as the ground truth.
'''
sdata = scipy.io.loadmat(os.path.join(dataPath, 'skin.mat'))['sdata']
ndata = scipy.io.loadmat(os.path.join(dataPath, 'nonskin.mat'))['ndata']
## Multivariate Gaussian for skin and non-skin
mvn_sskin = [MVND(sdata)]
mvn_nskin = [MVND(ndata)]
# Optain priors
mask = imageHelper()
mask.loadImageFromFile(os.path.join(dataPath, 'mask.png'))
# EXERCISE 2 - Compute the skin and nonskin prior
prior_skin, prior_nonskin = get_prior(mask)
print(prior_skin)
print("TRAINING DATA")
trainingmaskObj = imageHelper()
trainingmaskObj.loadImageFromFile(os.path.join(dataPath, 'mask.png'))
trainingimageObj = imageHelper()
trainingimageObj.loadImageFromFile(os.path.join(dataPath, 'image.png'))
# (img: imageHelper, mask: imageHelper, skin_mvnd: List[MVND], notSkin_mvnd: List[MVND], fig: str = "", prior_skin: float = 0.5, prior_nonskin: float = 0.5)
classify(trainingimageObj, trainingmaskObj, mvn_sskin, mvn_nskin,
"Training", prior_skin, prior_nonskin)
print("TEST DATA")
testmaskObj = imageHelper()
testmaskObj.loadImageFromFile(os.path.join(dataPath, 'mask-test.png'))
testimageObj = imageHelper()
testimageObj.loadImageFromFile(os.path.join(dataPath, 'test.png'))
classify(testimageObj, testmaskObj, mvn_sskin, mvn_nskin, "Test",
prior_skin, prior_nonskin)
plt.show()
def mvndSkinDetection() -> None:
'''
Skin detection - compute a MVND for both skin and non-skin.
Classify the test and training image using the classify helper function.
Note that the "mask" binary images are used as the ground truth.
'''
sdata = scipy.io.loadmat(os.path.join(dataPath, 'skin.mat'))['sdata']
ndata = scipy.io.loadmat(os.path.join(dataPath, 'nonskin.mat'))['ndata']
mvn_sskin = [MVND(sdata)]
mvn_nskin = [MVND(ndata)]
# Optain priors
mask = imageHelper()
mask.loadImageFromFile(os.path.join(dataPath, 'mask.png'))
# TODO: EXERCISE 2 - Compute the skin and nonskin prior -> solved
prior_skin, prior_nonskin = get_prior(mask)
print("TRAINING DATA")
trainingmaskObj = imageHelper()
trainingmaskObj.loadImageFromFile(os.path.join(dataPath, 'mask.png'))
trainingimageObj = imageHelper()
trainingimageObj.loadImageFromFile(os.path.join(dataPath, 'image.png'))
classify(trainingimageObj, trainingmaskObj, mvn_sskin, mvn_nskin,
"Training-MVND", prior_skin, prior_nonskin)
print("TEST DATA PORTRAIT")
testmaskObj = imageHelper()
testmaskObj.loadImageFromFile(os.path.join(dataPath, 'mask-test1.png'))
testimageObj = imageHelper()
testimageObj.loadImageFromFile(os.path.join(dataPath, 'test1.png'))
classify(testimageObj, testmaskObj, mvn_sskin, mvn_nskin,
"Test-portrait-MVND", prior_skin, prior_nonskin)
print("TEST DATA FAMILY")
testmaskObj = imageHelper()
testmaskObj.loadImageFromFile(os.path.join(dataPath, 'mask-test2.png'))
testimageObj = imageHelper()
testimageObj.loadImageFromFile(os.path.join(dataPath, 'test2.png'))
classify(testimageObj, testmaskObj, mvn_sskin, mvn_nskin,
"Test-family-MVND", prior_skin, prior_nonskin)
plt.show()