def train():
form = TrainForm()
if form.validate_on_submit():
classifier.train([form.data['example']], [form.data['label']])
flash('Thanks!', 'success')
else:
flash('An error occurred!', 'danger')
return redirect('/')
def train():
form = TrainForm()
if form.validate_on_submit():
classifier.train([form.data['example']], [form.data['label']])
flash('Thanks!', 'success')
else:
flash('An error occurred!', 'danger')
return redirect('/')
def trainui():
form = TrainForm()
if form.validate_on_submit():
print(request.form.to_dict())
args = copy.deepcopy(request.form.to_dict())
del args['csrf_token']
print(args['args'])
generate_yaml('train.yaml', 'test.yaml', args)
return redirect(url_for('trainImage', **args))
return render_template('train.html', title='nexo', form=form)
def home():
data1 = pd.read_csv('data.csv')
data1 = data1[data1.label != 'nu']
x = data1.iloc[:, 0].values
y = data1.iloc[:, 1].values
vectorizer = TfidfVectorizer(encoding='utf-8')
x = vectorizer.fit_transform(data1['text'])
x_train, x_test, y_train, y_test = train_test_split(x,
y,
test_size=0.25,
random_state=50)
model = svm.SVC(gamma='auto', probability=True)
model.fit(x_train, y_train)
x_score = model.score(x_train, y_train)
y_score = model.score(x_test, y_test)
sform = SearchForm()
# pos = 10
# neg = 18
if sform.validate_on_submit():
sent = sform.text.data
sent1 = np.array(list(sent))
se = vectorizer.transform(sent1)
pre = model.predict_proba(se)
pos = pre[0][0] * 100
neg = pre[0][1] * 100
flash(f'Training accuracy:{x_score}, and Testing accuracy: {y_score}',
'success')
return render_template('search.html', form=sform, pos=pos, neg=neg)
else:
return render_template('home.html', form=sform)
tform = TrainForm()
if tform.validate_on_submit():
flash('Training in progress', 'success')
return render_template('search.html', form='tform')
else:
flash('You should first train your model', 'danger')
def add_train():
if current_user.sostav != 'P':
return redirect('/')
else:
form = TrainForm()
if form.validate_on_submit():
session = db_session.create_session()
fdate = datetime.datetime.combine(form.date.data.date(),
form.time.data.time())
if session.query(Train).filter(
Train.date == fdate,
Train.sostav == form.sostav.data).first():
return render_template(
'add_train.html',
form=form,
message="У этого состава уже есть тренировка в это время")
train = Train(date=fdate,
sostav=form.sostav.data,
dur=int(form.dur.data),
user_id=int(current_user.id))
session.add(train)
session.commit()
return redirect('/success')
return render_template('add_train.html', form=form)
def home():
form = PredictForm()
if form.validate_on_submit():
cleaned = pattern.sub(' ', form.example.data.lower())
new_examples = [cleaned]
predictions, probs = classifier.predict(new_examples)
return render_template('result.html',
max_coef=classifier.get_max_coefficient(),
words=classifier.get_coefficients_for(
new_examples[0]),
example=new_examples[0],
pos_prob=probs[1],
neg_prob=probs[0],
form=TrainForm())
return render_template('index.html', form=form)
def index():
trainForm = TrainForm()
testForm = TestForm()
if trainForm.submitTrain.data and trainForm.validate_on_submit():
if request.method == "POST":
if os.path.exists('temp'):
directoryUtils.rmtree("temp")
os.mkdir('temp')
for f in trainForm.images.data:
# remove underscores to avoid ambiguity
# as secure_filename() replaces os.path.sep with underscore
safe_name = f.filename.replace('_', '')
fil = secure_filename(safe_name)
if allowed_file(fil, ['jpg', 'jpeg', 'png']):
f.save(os.path.join('temp', fil))
split = float(trainForm.split.data)
epochs = int(trainForm.epochs.data)
if os.path.isfile(LOG_FILE_TRAIN):
os.remove(LOG_FILE_TRAIN)
w = open(LOG_FILE_TRAIN, 'w+')
cmd = [
'python', '-u', 'modelTrain.py',
str(split),
str(epochs),
str(trainForm.width.data),
str(trainForm.height.data)
]
subprocess.Popen(cmd, stdout=w, stderr=subprocess.STDOUT)
w.close()
return redirect(url_for('trainResults'))
if testForm.submitTest.data and testForm.validate_on_submit():
if request.method == "POST":
if os.path.exists('temp'):
directoryUtils.rmtree("temp")
os.mkdir('temp')
os.mkdir(os.path.join("temp", "model"))
os.mkdir(os.path.join("temp", "label"))
os.mkdir(os.path.join("temp", "images"))
model = testForm.model.data
label = testForm.labels.data
if allowed_file(secure_filename(model.filename), ['h5']):
model.save(
os.path.join("temp", "model",
secure_filename(model.filename)))
else:
return 'Invalid model.'
if allowed_file(secure_filename(label.filename), ['dat']):
label.save(
os.path.join("temp", "label",
secure_filename(label.filename)))
else:
return 'Invalid labels.'
# process based on source
print('Beginning handover to prediction script.')
images_dir_path = None
if testForm.source.data == 'images':
for f in testForm.images.data:
if allowed_file(secure_filename(f.filename),
['jpeg', 'jpg', 'png']):
f.save(
os.path.join('temp', 'images',
secure_filename(f.filename)))
images_dir_path = 'temp'
elif testForm.source.data == 'path':
images_dir_path = testForm.path.data
if os.path.isfile(LOG_FILE_TEST):
os.remove(LOG_FILE_TEST)
w = open(LOG_FILE_TEST, 'w+')
cmd = ['python', '-u', 'modelPredict.py', images_dir_path]
subprocess.Popen(cmd, stdout=w, stderr=subprocess.STDOUT)
w.close()
return redirect(url_for('testResults'))
return render_template("HTML_Interface.html",
trainForm=trainForm,
testForm=testForm)
def train():
# persist model
# joblib.dump(clf, 'model.pkl')
cm = [[0, 0], [0, 0]]
score = 0
precision = 0
recall = 0
f1score = 0
form = TrainForm()
if request.method == 'POST':
# print(request.form['algorithm'])
# read breast cancer data set
breast_cancer_df = pd.read_csv('./static/breastcancer.csv')
# if breast_cancer_df[""]
# print(breast_cancer_df[breast_cancer_df["Class"]=='Benign'])
mapping = {'Malignant': 1, 'Benign': 0}
breast_cancer_df.replace({'Class': mapping}, inplace=True)
X = breast_cancer_df.iloc[:, :-1].values
y = breast_cancer_df.iloc[:, -1].values
mask = X[:, 5] == '?'
X[mask] = 'NaN'
imputer = Imputer(missing_values='NaN', strategy='mean', axis=0)
imputer = imputer.fit(X[:, 0:9])
X[:, 0:9] = imputer.transform(X[:, 0:9])
X_train, X_test, y_train, y_test = train_test_split(X,
y,
test_size=0.25,
random_state=0)
if request.form['algorithm'] == 'nb':
# # fit model
clf = GaussianNB()
elif request.form['algorithm'] == 'dt':
# # fit model
clf = DecisionTreeClassifier()
elif request.form['algorithm'] == 'rf':
# # fit model
clf = RandomForestClassifier()
# elif request.form['algorithm'] == 'svm':
# # # fit model
# clf = SVC()
# elif request.form['algorithm'] == 'logreg':
# # # fit model
# clf = LogisticRegression()
clf.fit(X_train, y_train)
# persist model
joblib.dump(clf, 'model.pkl')
y_pred = clf.predict(X_test)
cm = confusion_matrix(y_test, y_pred)
# print(classification_report(y_test,y_pred))
score = clf.score(X_test, y_test)
precision = precision_score(y_test, y_pred)
recall = recall_score(y_test, y_pred)
f1score = f1_score(y_test, y_pred)
# test = [[5,1,1,1,2,1,3,1,1],[8,10,10,8,7,10,9,7,1]]
# prediction = clf.predict_proba(test)
# print(prediction)
# flash(f'Model Created '+ str(request.form['algorithm']) + str(clf.score(X_test,y_test))+ str(cm),'success')
flash(f'Model Created ' + str(request.form['algorithm']), 'success')
# return redirect(url_for('train',cm = cm))
return render_template('train_model.html',
title='Train Model',
form=form,
cm=cm,
accuracy=score,
precision=precision,
recall=recall,
f1score=f1score)
return render_template('train_model.html',
title='Train Model',
form=form,
cm=cm,
accuracy=score,
precision=precision,
recall=recall,
f1score=f1score)
def index():
trainForm = TrainForm()
testForm = TestForm()
if trainForm.submit.data and trainForm.validate_on_submit():
if request.method == "POST":
if os.path.exists('temp'):
directoryUtils.rmtree("temp")
os.mkdir('temp')
for f in trainForm.images.data:
if allowed_file(secure_filename(f.filename),
['jpg', 'jpeg', 'png']):
fil = secure_filename(f.filename)
f.save(os.path.join('temp', fil))
split = float(trainForm.split.data)
epochs = int(trainForm.epochs.data)
print("Beginning handover to training script.")
modelTrain.trainModel(split, epochs)
modelList = []
query = re.compile(r'^(\w+)_(\d.\d{3})_(\d.\d{3}).h5$')
for file in os.listdir("output"):
matches = query.findall(file)
model = {}
if len(matches) > 0:
model['specialization'] = "Maximum Accuracy" if matches[0][
0] == "max_acc" else "Minimum Loss"
model['val_loss'] = '{:.3f}'.format(float(matches[0][1]))
model['val_acc'] = '{:.1f}%'.format(
float(matches[0][2]) * 100)
model['filename'] = file
modelList.append(model)
directoryUtils.rmtree("temp")
return render_template("trainResults.html", models=modelList)
if testForm.submit.data and testForm.validate_on_submit():
if request.method == "POST":
if os.path.exists('temp'):
directoryUtils.rmtree("temp")
os.mkdir('temp')
os.mkdir(os.path.join("temp", "model"))
os.mkdir(os.path.join("temp", "label"))
os.mkdir(os.path.join("temp", "images"))
model = testForm.model.data
label = testForm.labels.data
if allowed_file(secure_filename(model.filename), ['h5']):
model.save(
os.path.join("temp", "model",
secure_filename(model.filename)))
else:
return 'Invalid model.'
if allowed_file(secure_filename(label.filename), ['dat']):
label.save(
os.path.join("temp", "label",
secure_filename(label.filename)))
else:
return 'Invalid labels.'
for f in testForm.images.data:
if allowed_file(secure_filename(f.filename),
['jpeg', 'jpg', 'png']):
f.save(
os.path.join('temp', 'images',
secure_filename(f.filename)))
print("Beginning handover to prediction script.")
modelPredict.predict()
directoryUtils.rmtree("temp")
return render_template("testResults.html")
return render_template("HTML_Interface.html",
trainForm=trainForm,
testForm=testForm)