Feeding CNN outputs to a RNN

Dockerfile

@@ -0,0 +1,17 @@
+FROM debian:latest
+
+RUN apt-get -y update && apt-get install -y git python3-pip python3-dev python3-tk vim procps curl
+
+#Face classificarion dependencies & web application
+RUN pip3 install numpy scipy scikit-learn pillow tensorflow pandas h5py opencv-python==3.2.0.8 keras statistics pyyaml pyparsing cycler matplotlib Flask
+
+ADD . /ekholabs/face-classifier
+
+WORKDIR ekholabs/face-classifier
+
+ENV PYTHONPATH=$PYTHONPATH:src
+ENV FACE_CLASSIFIER_PORT=8084
+EXPOSE $FACE_CLASSIFIER_PORT
+
+ENTRYPOINT ["python3"]
+CMD ["src/web/faces.py"]

README.md

@@ -1,21 +1,51 @@
-# Face classification and detection
+# This repository is deprecated for at TF-2.0 rewrite visit:
+# https://github.com/oarriaga/paz
+------------------------------------------------
+# Face classification and detection.
 Real-time face detection and emotion/gender classification using fer2013/IMDB datasets with a keras CNN model and openCV.
 * IMDB gender classification test accuracy: 96%.
 * fer2013 emotion classification test accuracy: 66%.
 
+For more information please consult the [publication](https://github.com/oarriaga/face_classification/blob/master/report.pdf)
 
+# Emotion/gender examples:
+
+![alt tag](images/demo_results.png)
+
+Guided back-prop
+![alt tag](images/gradcam_results.png)
+
+Real-time demo:
 <div align='center'>
   <img src='images/color_demo.gif' width='400px'>
 </div>
 
-Emotion/gender classification of the [B-IT-BOTS](https://mas-group.inf.h-brs.de/?page_id=622) robotics team :) 
-![alt tag](images/predicted_test_image.png)
-
-Emotion examples
-![alt tag](images/emotion_classification.jpg)
+[B-IT-BOTS](https://mas-group.inf.h-brs.de/?page_id=622) robotics team :)
+![alt tag](images/robocup_team.png)
 
 ## Instructions
 
+### Run real-time emotion demo:
+> python3 video_emotion_color_demo.py
+
+### Run real-time guided back-prop demo:
+> python3 image_gradcam_demo.py
+
+### Make inference on single images:
+> python3 image_emotion_gender_demo.py <image_path>
+
+e.g.
+
+> python3 image_emotion_gender_demo.py ../images/test_image.jpg
+
+### Running with Docker
+
+With a few steps one can get its own face classification and detection running. Follow the commands below:
+
+* ```docker pull ekholabs/face-classifier```
+* ```docker run -d -p 8084:8084 --name=face-classifier ekholabs/face-classifier```
+* ```curl -v -F image=@[path_to_image]  http://localhost:8084/classifyImage > image.png```
+
 ### To train previous/new models for emotion classification:
 
 
@@ -38,7 +68,6 @@ Emotion examples
 * Untar the file:
 > tar -xfv imdb_crop.tar
 
-* Run the train_emotion_classification.py file
-> python3 train_emotion_classifier.py
-
+* Run the train_gender_classification.py file
+> python3 train_gender_classifier.py
 

REQUIREMENTS.txt

@@ -1,9 +1,7 @@
-keras==2.0.3
+keras==2.0.5
 tensorflow==1.1.0
 pandas==0.19.1
 numpy==1.12.1
 h5py==2.7.0
 statistics
-
-#optional for video_test.py
-opencv2-python==3.2.0
+opencv-python==3.2.0

src/models/cnn.py

@@ -10,6 +10,7 @@
 from keras import layers
 from keras.regularizers import l2
 
+
 def simple_CNN(input_shape, num_classes):
 
     model = Sequential()
@@ -48,11 +49,13 @@ def simple_CNN(input_shape, num_classes):
 
     model.add(Convolution2D(filters=256, kernel_size=(3, 3), padding='same'))
     model.add(BatchNormalization())
-    model.add(Convolution2D(filters=num_classes, kernel_size=(3, 3), padding='same'))
+    model.add(Convolution2D(
+        filters=num_classes, kernel_size=(3, 3), padding='same'))
     model.add(GlobalAveragePooling2D())
-    model.add(Activation('softmax',name='predictions'))
+    model.add(Activation('softmax', name='predictions'))
     return model
 
+
 def simpler_CNN(input_shape, num_classes):
 
     model = Sequential()
@@ -100,21 +103,22 @@ def simpler_CNN(input_shape, num_classes):
                             strides=(2, 2), padding='same'))
 
     model.add(Flatten())
-    #model.add(GlobalAveragePooling2D())
-    model.add(Activation('softmax',name='predictions'))
+    # model.add(GlobalAveragePooling2D())
+    model.add(Activation('softmax', name='predictions'))
     return model
 
+
 def tiny_XCEPTION(input_shape, num_classes, l2_regularization=0.01):
     regularization = l2(l2_regularization)
 
     # base
     img_input = Input(input_shape)
     x = Conv2D(5, (3, 3), strides=(1, 1), kernel_regularizer=regularization,
-                                            use_bias=False)(img_input)
+               use_bias=False)(img_input)
     x = BatchNormalization()(x)
     x = Activation('relu')(x)
     x = Conv2D(5, (3, 3), strides=(1, 1), kernel_regularizer=regularization,
-                                            use_bias=False)(x)
+               use_bias=False)(x)
     x = BatchNormalization()(x)
     x = Activation('relu')(x)
 
@@ -191,10 +195,10 @@ def tiny_XCEPTION(input_shape, num_classes, l2_regularization=0.01):
     x = layers.add([x, residual])
 
     x = Conv2D(num_classes, (3, 3),
-            #kernel_regularizer=regularization,
-            padding='same')(x)
+               # kernel_regularizer=regularization,
+               padding='same')(x)
     x = GlobalAveragePooling2D()(x)
-    output = Activation('softmax',name='predictions')(x)
+    output = Activation('softmax', name='predictions')(x)
 
     model = Model(img_input, output)
     return model
@@ -206,11 +210,11 @@ def mini_XCEPTION(input_shape, num_classes, l2_regularization=0.01):
     # base
     img_input = Input(input_shape)
     x = Conv2D(8, (3, 3), strides=(1, 1), kernel_regularizer=regularization,
-                                            use_bias=False)(img_input)
+               use_bias=False)(img_input)
     x = BatchNormalization()(x)
     x = Activation('relu')(x)
     x = Conv2D(8, (3, 3), strides=(1, 1), kernel_regularizer=regularization,
-                                            use_bias=False)(x)
+               use_bias=False)(x)
     x = BatchNormalization()(x)
     x = Activation('relu')(x)
 
@@ -287,14 +291,15 @@ def mini_XCEPTION(input_shape, num_classes, l2_regularization=0.01):
     x = layers.add([x, residual])
 
     x = Conv2D(num_classes, (3, 3),
-            #kernel_regularizer=regularization,
-            padding='same')(x)
+               # kernel_regularizer=regularization,
+               padding='same')(x)
     x = GlobalAveragePooling2D()(x)
-    output = Activation('softmax',name='predictions')(x)
+    output = Activation('softmax', name='predictions')(x)
 
     model = Model(img_input, output)
     return model
 
+
 def big_XCEPTION(input_shape, num_classes):
     img_input = Input(input_shape)
     x = Conv2D(32, (3, 3), strides=(2, 2), use_bias=False)(img_input)
@@ -331,10 +336,10 @@ def big_XCEPTION(input_shape, num_classes):
     x = MaxPooling2D((3, 3), strides=(2, 2), padding='same')(x)
     x = layers.add([x, residual])
     x = Conv2D(num_classes, (3, 3),
-            #kernel_regularizer=regularization,
-            padding='same')(x)
+               # kernel_regularizer=regularization,
+               padding='same')(x)
     x = GlobalAveragePooling2D()(x)
-    output = Activation('softmax',name='predictions')(x)
+    output = Activation('softmax', name='predictions')(x)
 
     model = Model(img_input, output)
     return model
@@ -343,11 +348,11 @@ def big_XCEPTION(input_shape, num_classes):
 if __name__ == "__main__":
     input_shape = (64, 64, 1)
     num_classes = 7
-    #model = tiny_XCEPTION(input_shape, num_classes)
-    #model.summary()
-    #model = mini_XCEPTION(input_shape, num_classes)
-    #model.summary()
-    #model = big_XCEPTION(input_shape, num_classes)
-    #model.summary()
+    # model = tiny_XCEPTION(input_shape, num_classes)
+    # model.summary()
+    # model = mini_XCEPTION(input_shape, num_classes)
+    # model.summary()
+    # model = big_XCEPTION(input_shape, num_classes)
+    # model.summary()
     model = simple_CNN((48, 48, 1), num_classes)
     model.summary()