Merge pull request #12 from alessiosavi/v0.3.0

V0.3.0

README.md

@@ -347,6 +347,8 @@ The tool is powered with `Flask_MonitoringDashboard` that expose some useful uti
 - [Pillow](https://github.com/python-pillow/Pillow) The friendly PIL fork (Python Imaging Library)
 - [scikit-learn](https://github.com/scikit-learn/scikit-learn) Machine learning in Python
 
+***NOTE***: If you encounter an error during `pip install -r requirements.txt`, it's possible that you have not installed `cmake`. `dlib` need `cmake`, and you can install using `apt install cmake -y` (debian/ubuntu) or `yum install cmake -y` (CentOS/Fedora/RedHat).
+
 ## Table Of Contents
 
 - [PyRecognizer](#pyrecognizer)

api/Api.py

@@ -16,7 +16,7 @@
 
 
 
-def predict_image(img_path, clf, PREDICTION_PATH):
+def predict_image(img_path, clf, PREDICTION_PATH,treshold=45):
     """
 
     :param PREDICTION_PATH: global variable where image recognized are saved
@@ -30,7 +30,7 @@ def predict_image(img_path, clf, PREDICTION_PATH):
         prediction = None
     else:
         log.debug("predict_image | Predicting {}".format(img_path))
-        prediction = clf.predict(img_path)
+        prediction = clf.predict(img_path,treshold)
         log.debug("predict_image | Result: {}".format(prediction))
 
     # Manage error
@@ -132,3 +132,6 @@ def tune_network(folder_uncompress, zip_file, clf):
             response.description))
 
         return response.__dict__
+
+
+

api/templates/upload.html

@@ -10,6 +10,9 @@ <h1>Face recognition tool</h1>
 <form enctype=multipart/form-data method=post>
     <input name=file type=file>
     <input type=submit value=Upload>
+    <br><br>
+    <label for="treshold">Treshold</label>
+    <input type="text" name="treshold"  value="45">
     <input name=_csrf_token type=hidden value="{{ csrf_token() }}">
 </form>
 {% with messages = get_flashed_messages() %}

conf/test.json

@@ -23,7 +23,7 @@
   "classifier": {
     "trainin_dir": "dataset/images/",
     "model_path": "dataset/model/",
-    "timestamp": "20191125_004201"
+    "timestamp": "20191202_153034"
   },
   "data": {
     "test_data": "/tmp/test_data/"

dataset/model/20191202_153034/model.json

@@ -0,0 +1 @@
+{"classifier_file": "20191202_153034/model.clf", "params": {"activation": "identity", "hidden_layer_sizes": [128], "learning_rate": "constant", "solver": "adam"}}

datastructure/Administrator.py

@@ -74,7 +74,7 @@ def add_user(self) -> bool:
             return False
 
         log.warning("Encrypting password -> {}".format(self.password))
-        self.password = self.encrypt_password(self.password)
+        self.password = self.encrypt_password(str(self.password))
         log.warning("Passwrod encrypted {}".format(self.password))
         self.redis_client.set(self.get_name(), self.password)
         log.warning("User {} registered!".format(self.get_name()))
@@ -99,7 +99,7 @@ def get_name(self) -> str:
     def encrypt_password(self, plain_text_password: str) -> str:
         # Hash a password for the first time
         #   (Using bcrypt, the salt is saved into the hash itself)
-        return bcrypt.hashpw(plain_text_password.encode('utf-8'), bcrypt.gensalt())
+        return bcrypt.hashpw(plain_text_password, bcrypt.gensalt())
 
     def check_password(self, plain_text_password: str, hashed_password: str) -> bool:
         # Check hashed password. Using bcrypt, the salt is saved into the hash itself

datastructure/Classifier.py

@@ -120,6 +120,7 @@ def train(self, X, Y, timestamp):
 
         X_train, x_test, Y_train, y_test = train_test_split(
             X, Y, test_size=0.25)
+
         log.debug("train | Training ...")
         self.classifier.fit(X_train, Y_train)
         log.debug("train | Model Trained!")
@@ -144,18 +145,18 @@ def tuning(self, X, Y, timestamp):
 
         X_train, x_test, Y_train, y_test = train_test_split(
             X, Y, test_size=0.25)
-        self.classifier = MLPClassifier(max_iter=200)
+        self.classifier = MLPClassifier(max_iter=250)
         # Hyperparameter of the neural network (MLP) to tune
+        # Faces are encoded using 128 points
         parameter_space = {
-            'hidden_layer_sizes': [(200,), (100, 200, 100), (100,), ],
-            # 'activation': ['identity', 'tanh', 'relu'],
-            'activation': ['identity'],
+            'hidden_layer_sizes': [(128,), (200,), (200, 128,), ],
+            'activation': ['identity', 'tanh', 'relu'],
             'solver': ['adam'],
-            'learning_rate': ['constant'],
+            'learning_rate': ['constant', 'adaptive'],
         }
         log.debug("tuning | Parameter -> {}".format(pformat(parameter_space)))
         grid = GridSearchCV(self.classifier, parameter_space,
-                            cv=2, scoring='accuracy', verbose=20, n_jobs=None, pre_dispatch=1)
+                            cv=2, scoring='accuracy', verbose=20, n_jobs=8)
         grid.fit(X_train, Y_train)
         log.info("TUNING COMPLETE | DUMPING DATA!")
         # log.info("tuning | Grid Scores: {}".format(pformat(grid.grid_scores_)))

main.py

@@ -14,6 +14,7 @@
 from api.Api import predict_image, train_network, tune_network
 from datastructure.Classifier import Classifier
 from datastructure.Administrator import Administrator
+from datastructure.Response import Response
 from utils.util import init_main_data, random_string, secure_request
 
 # ===== LOAD CONFIGURATION FILE =====
@@ -74,10 +75,33 @@ def predict():
         log.warning("predict_api | No file choosed!")
         return redirect(request.url)  # Return to HTML page [GET]
     file = request.files['file']
+    treshold = request.form.get('treshold')
+    log.debug("Recived file [{}] and treshold [{}]".format(file,treshold))
+    try:
+        treshold = int(treshold)
+    except ValueError:
+        log.error("Unable to convert treshold")
+        response = Response()
+        response.error = "UNABLE_CAST_INT"
+        response.description = "Treshold is not an integer!"
+        response.status = "KO"
+        return jsonify(response=response.__dict__)
+    if not 0 <= treshold <= 100:
+        log.error("Treshold wrong value")
+        response = Response()
+        response.error = "TRESHOLD_ERROR_VALUE"
+        response.description = "Treshold have to be greater than 0 and lesser than 100!"
+        response.status = "KO"
+        return jsonify(response=response.__dict__)
+
+    treshold /= 100
+
+
+    log.debug("Recived file {} and treshold {}".format(file,treshold))
     filename = secure_filename(file.filename)
     img_path = os.path.join(app.config['UPLOAD_FOLDER'], filename)
     file.save(img_path)
-    return jsonify(response=predict_image(img_path, clf, TMP_UPLOAD_PREDICTION))
+    return jsonify(response=predict_image(img_path, clf, TMP_UPLOAD_PREDICTION,treshold))
 
 
 @app.route('/train', methods=['GET'])

requirements.txt

@@ -1,10 +1,11 @@
-Flask==1.1.1
+Flask_Login==0.2.9
 face_recognition==1.2.3
-scikit-learn==0.21.3
-tqdm==4.39.0
+redis==3.3.11
 Flask_MonitoringDashboard==3.0.5
-werkzeug==0.16.0
+Flask==1.1.1
+numpy==1.17.4
+tqdm==4.39.0
+Werkzeug==0.16.0
 Pillow==6.2.1
-numpy
-redis
-
+py-bcrypt
+scikit_learn==0.21.3