def GetQuestionMyOneDirection(anchor_ims, checkpoint_dir, data_bin, data_path,
n_features, direction):
print(anchor_ims)
dir_ = "/Users/admin/Desktop/UVA/third/CycleGAN-tensorflow/"
for i in glob(os.path.join("result/", "*.jpg")):
subprocess.call(["rm", i])
for anchor_im in anchor_ims:
subprocess.call(["cp", data_path[anchor_im], "result/"])
for i in glob(os.path.join(dir_ + "test1/", "*.jpg")):
subprocess.call(["rm", i])
#for i in glob(os.path.join(dir_ + "test/", "*.jpg")):
# subprocess.call(["rm", i])
subprocess.call([
"python", dir_ + "main.py", "--dataset_a", "result/", "--dataset_b",
"result/", "--which_direction", direction, "--phase", "test",
"--checkpoint_dir", checkpoint_dir, "--test_dir", dir_ + "test/",
"--test_dir_a", dir_ + "test1/", "--test_dir_b", dir_ + "test/"
])
images = []
file = glob(os.path.join(dir_ + "test1/", "*.jpg"))
for f in file:
gist = Gist(f)
svm = []
for i in range(n_features):
svm.append(
np.genfromtxt("../CycleGAN_shoes/Black/" + str(i + 1) +
".txt"))
res = np.array([np.dot(svm[i], gist) for i in range(n_features)])
print(res)
images += FindTheClosestImage(data_bin.keys(), data_bin, res, 0, 1,
n_features)
return images
def main():
args = parse_args()
path = "../data/lgs/" + args.lg + "/" + args.lg
features, by_speaker, no_normalize, to_normalize = get_features(
args.features_csv, args.lg)
x, y, data = read_norm(args.lg, by_speaker, no_normalize, to_normalize,
features)
svm = []
rf = []
for feat in args.to_ablate.split(", "):
# Remove feature from x
x_abl = x.drop([feat], axis=1, inplace=False)
# Run SVM and RF
SVM_rs_aprf = SVM_rs(x_abl, y, features, args.lg)
SVM_rs_aprf.insert(0, feat)
RF_rs_aprf = RF_rs(x_abl, y, features, args.lg)
RF_rs_aprf.insert(0, feat)
svm.append(SVM_rs_aprf)
rf.append(RF_rs_aprf)
svm = pd.DataFrame(svm,
columns=[
'feat', 'acc', 'pB', 'pM', 'pC', 'rB', 'rM', 'rC',
'fB', 'fM', 'fC'
])
rf = pd.DataFrame(rf,
columns=[
'feat', 'acc', 'pB', 'pM', 'pC', 'rB', 'rM', 'rC',
'fB', 'fM', 'fC'
])
svm.to_csv(path + "-abl-SVM.csv")
rf.to_csv(path + "-abl-RF.csv")
return
def main():
args = parse_args()
path = "../data/lgs/" + args.lg + "/" + args.lg
features, by_speaker, no_normalize, to_normalize, cat_dict = get_features(args.features_csv, args.lg)
x, y, data = read_norm(args.lg, by_speaker, no_normalize, to_normalize, features)
svm = []
rf = []
for cat in cat_dict:
x_abl = x
# Loop through features in each cat and remove them
for feat in cat_dict[cat]:
x_abl = x_abl.drop([feat], axis = 1, inplace = False)
# Run classifiers
SVM_rs_aprf = SVM_rs(x_abl, y, features, args.lg)
SVM_rs_aprf.insert(0, cat)
SVM_rs_aprf.append(len(cat_dict[cat]))
RF_rs_aprf = RF_rs(x_abl, y, features, args.lg)
RF_rs_aprf.insert(0, cat)
RF_rs_aprf.append(len(cat_dict[cat]))
svm.append(SVM_rs_aprf)
rf.append(RF_rs_aprf)
svm = pd.DataFrame(svm, columns = ['cat', 'acc', 'f1', 'pB', 'pM', 'pC', 'rB', 'rM', 'rC', 'fB', 'fM', 'fC', 'n feat'])
rf = pd.DataFrame(rf, columns = ['cat', 'acc', 'f1', 'pB', 'pM', 'pC', 'rB', 'rM', 'rC', 'fB', 'fM', 'fC', 'n feat'])
svm.to_csv(path + "-abl-SVM.csv")
rf.to_csv(path + "-abl-RF.csv")
"""
def pick_clf(labels, features):
svm = []
dt = []
knn = []
for i in range(100):
svm.append(test_svm(labels, features))
dt.append(test_dt(labels, features))
knn.append(test_knn(labels, features))
svm = sum([x[0] for x in svm]) / len(svm)
dt = sum([x[0] for x in dt]) / len(dt)
knn = sum([x[0] for x in knn]) / len(knn)
print("average accuracy using svm: ", round(svm, 2))
print("average accuracy using dt: ", round(dt, 2))
print("average accuracy using knn: ", round(knn, 2))
def main():
np.set_printoptions(suppress=True)
data, target = loadCsv('data2.csv')
#statistics(data)
dataset_clean, target_clean = distanceMatOut(np.array(data),
np.array(target))
bayes_simple = []
logreg = []
bayes_quad = []
percep = []
mlp = []
svm = []
for i in range(0, 10):
data_train, data_test, target_train, target_test = kfoldSplit(
data, target, 10, i)
#data_train = zscore(data_train)
print "Bayesian Classifier"
stats = bayesianClassifier(data_train, data_test, target_train,
target_test)
bayes_simple.append(stats)
print "LogisticRegression"
stats = logisticRegression(data_train, data_test, target_train,
target_test)
logreg.append(stats)
print "Quadratic Bayesian Classifier"
stats = quadraticClassifier(data_train, data_test, target_train,
target_test)
bayes_quad.append(stats)
#data_test = zscore(data_test)
scaler = StandardScaler()
scaler.fit(data_train)
data_train = scaler.transform(data_train)
data_test = scaler.transform(data_test)
print "Perceptron"
stats = perceptron(data_train, data_test, target_train, target_test)
percep.append(stats)
print "MLP"
stats = multiLayerPerceptron(data_train, data_test, target_train,
target_test)
mlp.append(stats)
print "SVM"
stats = svmLinearKernel(data_train, data_test, target_train,
target_test)
print stats
svm.append(stats)
print "Bayesian Classifier"
conf_bayes = np.mean(np.transpose(np.array(bayes_simple)), axis=1)
base_name = 'naive_bayes'
plotPREREC(conf_bayes, '../img/' + base_name + '_rec.png')
saveConfMat(conf_bayes, base_name + '_conf.csv')
print conf_bayes
print "LogisticRegression"
conf_logreg = np.mean(np.transpose(np.array(logreg)), axis=1)
base_name = 'log_reg'
plotPREREC(conf_logreg, '../img/' + base_name + '_rec.png')
saveConfMat(conf_logreg, base_name + '_conf.csv')
print conf_logreg
print "Quadratic Bayesian Classifier"
conf_quad_bayes = np.mean(np.transpose(np.array(bayes_quad)), axis=1)
print conf_quad_bayes
base_name = 'quad_bayes'
plotPREREC(conf_quad_bayes, '../img/' + base_name + '_rec.png')
saveConfMat(conf_quad_bayes, base_name + '_conf.csv')
print "Perceptron"
conf_perc = np.mean(np.transpose(np.array(percep)), axis=1)
print conf_perc
base_name = 'perc'
plotPREREC(conf_perc, '../img/' + base_name + '_rec.png')
saveConfMat(conf_perc, base_name + '_conf.csv')
print "MLP"
conf_mlp = np.mean(np.transpose(np.array(mlp)), axis=1)
print conf_mlp
base_name = 'mlp'
plotPREREC(conf_mlp, '../img/' + base_name + '_rec.png')
saveConfMat(conf_mlp, base_name + '_conf.csv')
print "SVM"
conf_svm = np.mean(np.transpose(np.array(svm)), axis=1)
print conf_svm
base_name = 'svm'
plotPREREC(conf_svm, '../img/' + base_name + '_rec.png')
saveConfMat(conf_svm, base_name + '_conf.csv')
#print ("SVM:Train set score predict: {:.6f}".format(accuracy_svc_train))
print("SVM:Test set score: {:.6f}".format(linearSVC.score(X_test, y_test)))
#preds_test_svc = linearSVC.predict(X_test)
#accuracy_svc_test = accuracy_score(y_test,preds_test_svc)
#print ("SVM:Test set score predict: {:.6f}".format(accuracy_svc_test))
return linearSVC.score(X_test, y_test)
# In[14]:
svm = []
for i in range(1, iterations):
accuracy_svc = svm_30_features()
#print("accuracy_svm=%f "%accuracy_svm)
svm.append(accuracy_svc)
print(svm)
# ## SVM regression on PCA reduced set
# In[15]:
#svc = svm.SVC(C=10)
#svc.fit(train,train_labels)
#preds_svc_reduced = svc.predict(test)
#accuracy_svc_reduced = accuracy_score(test_labels,preds_svc_reduced)
#print(train.shape)
#print("accuracy_svm_reduced=%f "%accuracy_svc_reduced)
def svm_30_features_reduced():