feature(Gaze estimation):

src/face_detection.py

@@ -7,7 +7,7 @@
 import cv2
 from openvino.inference_engine import IENetwork, IECore
 
-class Model_X:
+class FaceDetector:
     '''
     Class for the Face Detection Model.
     '''
@@ -73,7 +73,6 @@ def preprocess_output(self, outputs, image):
         height = image.shape[0]
         width = image.shape[1]
         faces = []
-        count = 1
         for box in outputs[0][0]:
             if box[2] >= 0.6:
                 x_min = int(box[3] * width)
@@ -104,7 +103,7 @@ def main():
     CPU_EXTENSION_MAC = '/opt/intel/openvino_2019.3.376/deployment_tools/inference_engine/lib/intel64/libcpu_extension.dylib'
     model_name = 'models/intel/face-detection-adas-binary-0001/INT1/face-detection-adas-binary-0001'
     image = 'bin/test_image2.png'
-    model = Model_X(model_name=model_name, device='CPU', extensions=CPU_EXTENSION_MAC)
+    model = FaceDetector(model_name=model_name, device='CPU', extensions=CPU_EXTENSION_MAC)
     model.load_model()
     image = cv2.imread(image)
     pred = model.predict(image)

src/facial_landmarks_detection.py

@@ -7,7 +7,7 @@
 import cv2
 from openvino.inference_engine import IENetwork, IECore
 
-class Model_X:
+class FacialLandmarksDetector:
     '''
     Class for the Face Detection Model.
     '''
@@ -81,12 +81,12 @@ def get_eyes(self, coords, image):
         for key in coords.keys():
             x_coord = coords[key][0]
             y_coord = coords[key][1]
-            x_min = int(x_coord * width - 30)
-            y_min = int(y_coord * height - 30)
-            x_max = int(x_coord * width + 30)
-            y_max = int(y_coord * height + 30)
-            croped_image = image[y_min:y_max, x_min:x_max]
-            eyes[key] = croped_image
+            x_min = int(x_coord * (width - 30))
+            y_min = int(y_coord * (height - 30))
+            x_max = int(x_coord * (width + 30))
+            y_max = int(y_coord * (height + 30))
+            cropped_image = image[y_min:y_max, x_min:x_max]
+            eyes[key] = cropped_image
         return eyes
 
 def draw_landmarks(coords, frame, image):
@@ -113,7 +113,7 @@ def main():
     CPU_EXTENSION_MAC = '/opt/intel/openvino_2019.3.376/deployment_tools/inference_engine/lib/intel64/libcpu_extension.dylib'
     model_name = 'models/intel/landmarks-regression-retail-0009/FP16/landmarks-regression-retail-0009'
     image = 'bin/test_image2.png'
-    model = Model_X(model_name=model_name, device='CPU', extensions=CPU_EXTENSION_MAC)
+    model = FacialLandmarksDetector(model_name=model_name, device='CPU', extensions=CPU_EXTENSION_MAC)
     model.load_model()
     image = cv2.imread(image)
     pred, frame = model.predict(image)

src/gaze_estimation.py

@@ -0,0 +1,141 @@
+'''
+This is a sample class for a model. You may choose to use it as-is or make any changes to it.
+This has been provided just to give you an idea of how to structure your model class.
+'''
+import os
+import sys
+import cv2
+import numpy as np
+from openvino.inference_engine import IENetwork, IECore
+
+# import models
+from head_pose_estimation import HeadPoseEstimation
+from face_detection import FaceDetector
+from facial_landmarks_detection import FacialLandmarksDetector
+
+from draw_image import draw_boxes
+
+class GazeEstimation:
+    '''
+    Class for the Face Detection Model.
+    '''
+    def __init__(self, model_name, device='CPU', extensions=None):
+        '''
+        TODO: Use this to set your instance variables.
+        '''
+        self.model_name = model_name
+        self.device = device
+        self.extensions = extensions
+
+    def load_model(self):
+        '''
+        TODO: You will need to complete this method.
+        This method is for loading the model to the device specified by the user.
+        If your model requires any Plugins, this is where you can load them.
+        '''
+        model_bin = self.model_name + '.bin'
+        model_xml = self.model_name + '.xml'
+        self.model = IENetwork(model_xml, model_bin)
+        self.plugin = IECore()
+        if self.extensions:
+            self.plugin.add_extension(self.extensions, self.device)
+        self.exec_net = self.plugin.load_network(self.model, device_name=self.device, num_requests=1)
+        self.check_model()
+
+    def predict(self, inputs):
+        '''
+        TODO: You will need to complete this method.
+        This method is meant for running predictions on the input image.
+        '''
+        inputs = self.preprocess_input(inputs)
+        self.exec_net.start_async(request_id=0, inputs=inputs)
+        if self.exec_net.requests[0].wait(-1) == 0:
+            return self.exec_net.requests[0].outputs[self.output_name]
+
+    def check_model(self):
+        self.input_names = list(self.model.inputs.keys())
+        self.input_shapes = {}
+        for name in self.input_names:
+           self.input_shapes[name] = self.model.inputs[name].shape
+
+        self.output_name = next(iter(self.model.outputs))
+        self.output_shape = self.model.outputs[self.output_name].shape
+
+    def preprocess_input(self, inputs):
+        '''
+        Before feeding the data into the model for inference,
+        you might have to preprocess it. This function is where you can do that.
+        '''
+        left_eye_image = inputs['left_eye_image']
+        right_eye_image = inputs['right_eye_image']
+        head_pose = inputs['head_pose_angles']
+        print(left_eye_image.shape, right_eye_image.shape, head_pose.shape)
+        # shape for left and right eyes are the same
+        # using the shape of left eye
+        height = self.input_shapes['left_eye_image'][2]
+        width = self.input_shapes['left_eye_image'][3]
+        left_eye_image = cv2.resize(left_eye_image, (width, height))
+        left_eye_image = np.expand_dims(left_eye_image.transpose((2, 0, 1)), axis=0)
+
+        right_eye_image = cv2.resize(right_eye_image, (width, height))
+        right_eye_image = np.expand_dims(right_eye_image.transpose((2, 0, 1)), axis=0)
+
+        inputs['left_eye_image'] = left_eye_image
+        inputs['right_eye_image'] = right_eye_image
+        inputs['head_pose_angles'] = head_pose
+
+        return inputs
+
+    def preprocess_output(self, outputs):
+        '''
+        Before feeding the output of this model to the next model,
+        you might have to preprocess the output. This function is where you can do that.
+        '''
+        raise NotImplementedError
+
+def main():
+    CPU_EXTENSION_MAC = '/opt/intel/openvino_2019.3.376/deployment_tools/inference_engine/lib/intel64/libcpu_extension.dylib'
+    gaze_model = 'models/intel/gaze-estimation-adas-0002/FP16/gaze-estimation-adas-0002'
+    face_detector_model = 'models/intel/face-detection-adas-binary-0001/INT1/face-detection-adas-binary-0001'
+    facial_landmark_model = 'models/intel/landmarks-regression-retail-0009/FP16/landmarks-regression-retail-0009'
+    head_pose_model = 'models/intel/head-pose-estimation-adas-0001/FP16/head-pose-estimation-adas-0001'
+
+    image = 'bin/test-image1.jpg'
+
+    # Initialize the models
+    face_detector = FaceDetector(model_name=face_detector_model, device='CPU', extensions=CPU_EXTENSION_MAC)
+    facial_landmarks = FacialLandmarksDetector(model_name=facial_landmark_model, device='CPU', extensions=CPU_EXTENSION_MAC)
+    head_pose_estimation = HeadPoseEstimation(model_name=head_pose_model, device='CPU', extensions=CPU_EXTENSION_MAC)
+    gaze_estimation = GazeEstimation(model_name=gaze_model, device='CPU', extensions=CPU_EXTENSION_MAC)
+
+    # Load the models
+    face_detector.load_model()
+    facial_landmarks.load_model()
+    head_pose_estimation.load_model()
+    gaze_estimation.load_model()
+
+    image = cv2.imread(image)
+    pred = face_detector.predict(image)
+    image = face_detector.preprocess_output(pred, image)
+
+
+    head_pose = head_pose_estimation.predict(image[0])
+    head_pose = np.array([head_pose])
+
+    landmarks = facial_landmarks.predict(image[0])
+    eyes_coords = facial_landmarks.preprocess_output(landmarks[0])
+    eyes = facial_landmarks.get_eyes(eyes_coords, image[0])
+    left_eye_image = eyes['left_eye']
+    right_eye_image = eyes['right_eye']
+    cv2.imwrite('new_left_eyes.jpg', left_eye_image)
+    cv2.imwrite('new_right_eyes.jpg', right_eye_image)
+
+    gaze_estimate = gaze_estimation.predict({
+        'left_eye_image': eyes['left_eye'],
+        'right_eye_image': eyes['right_eye'],
+        'head_pose_angles': head_pose})
+
+    print(gaze_estimate)
+
+if __name__ == '__main__':
+    main()

src/head_pose_estimation.py

@@ -8,7 +8,7 @@
 import numpy as np
 from openvino.inference_engine import IENetwork, IECore
 
-class Model_X:
+class HeadPoseEstimation:
     '''
     Class for the Face Detection Model.
     '''
@@ -79,7 +79,7 @@ def main():
     CPU_EXTENSION_MAC = '/opt/intel/openvino_2019.3.376/deployment_tools/inference_engine/lib/intel64/libcpu_extension.dylib'
     model_name = 'models/intel/head-pose-estimation-adas-0001/FP16/head-pose-estimation-adas-0001'
     image = 'bin/test-image1.jpg'
-    model = Model_X(model_name=model_name, device='CPU', extensions=CPU_EXTENSION_MAC)
+    model = HeadPoseEstimation(model_name=model_name, device='CPU', extensions=CPU_EXTENSION_MAC)
     model.load_model()
     image = cv2.imread(image)
     pred = model.predict(image)