changed ReadME.md and get_face_recognition in order to make it work w…

…ith facenet model

image_recognition_face_recognition/README.md

@@ -1,34 +1,32 @@
 # image_recognition_face_recognition
 
-NEEDS WORK
+TO DO
 
-Face recognition with use of Openpose (https://cmusatyalab.github.io/openface/)
+Face recognition with use of Facenet (https://github.com/timesler/facenet-pytorch/tree/master)
+Paper can be found: [paper](https://arxiv.org/pdf/1503.03832.pdf)
 
 ## Installation
 
 See https://github.com/tue-robotics/image_recognition
 
-Make sure that openface is correctly installed. Installation instructions can be found here: https://cmusatyalab.github.io/openface/setup/. Make sure you have installed **CUDA8**. CUDA10 is not working: https://github.com/torch/torch7/issues/1180
+Make sure you have installed **CUDA8**.
 
-```bash
-export TORCH_INSTALL_PATH=~/torch
-export OPENFACE_INSTALL_PATH=~/openface
-git clone https://github.com/torch/distro.git ~/torch --recursive && \
-cd $TORCH_INSTALL_PATH && bash install-deps && ./install.sh && \
-for NAME in dpnn nn optim optnet csvigo cutorch cunn fblualib torchx tds; do $TORCH_INSTALL_PATH/install/bin/luarocks install $NAME; done && \
-sudo -H pip install dlib && \
-git clone https://github.com/cmusatyalab/openface.git $OPENFACE_INSTALL_PATH && \
-cd $OPENFACE_INSTALL_PATH && sudo -H python setup.py install && \
-./models/get-models.sh
-```
+bash commends will be available when merged
 
 ## How-to
 
+### Run Face Recognition Node
+
+Run the command why using your camera as an input (use roscamera node):
+
+    rosrun cv_camera cv_camera_node
+    rosrun image_recognition_face_recognition face_recognition_node image:=/cv_camera/image_raw 
+
 ### ROS Node
 
-Run the image_recognition_openface node in one terminal (Specify the dlib and openface_net path as ROS parameter):
+Run the image_recognition_face_detection node in one terminal (Specify the dlib and openface_net path as ROS parameter):
 
-    rosrun image_recognition_openface face_recognition_node
+    rosrun image_recognition_face_detection face_recognition_node
 
 Run the rqt annotation client (https://github.com/tue-robotics/image_recognition_rqt)
 
@@ -58,6 +56,7 @@ Again configure the service you want to call with the gear-wheel in the top-righ
 
 You will see that the result of the detection will prompt in a dialog combo box. Also the detections will be drawn on the image.
 
+
 ### Command line
 
 Command line interface to test the detection / recognition based on an image:
@@ -66,14 +65,6 @@ Command line interface to test the detection / recognition based on an image:
 
 Run the command on an example image:
 
-    rosrun image_recognition_openface get_face_recognition `rospack find image_recognition_openface`/doc/example.png
-
-This will lookup this image in the image_recognition_openface/doc folder and perform recognitions
-
-![Example](doc/example.png)
-
-Output:
-
-    [RecognizedFace(roi=(374, 188, 108, 123), l2_distances=[]), RecognizedFace(roi=(72, 147, 88, 105), l2_distances=[]), RecognizedFace(roi=(377, 95, 74, 86), l2_distances=[]), RecognizedFace(roi=(149, 26, 74, 86), l2_distances=[]), RecognizedFace(roi=(52, 47, 75, 86), l2_distances=[]), RecognizedFace(roi=(246, 115, 88, 102), l2_distances=[]), RecognizedFace(roi=(0, 0, 42, 60), l2_distances=[]), RecognizedFace(roi=(336, 33, 74, 86), l2_distances=[]), RecognizedFace(roi=(228, 0, 62, 60), l2_distances=[])]
+    rosrun image_recognition_face_recognition get_face_recognition `rospack find image_recognition_face_recognition`/doc/example.png
 
-Since no faces were trained, the l2_distances will not be calculated of-course.
+![Example](doc/example.png)

image_recognition_face_recognition/scripts/get_face_recognition

@@ -5,8 +5,9 @@ import argparse
 
 import cv2
 from image_recognition_msgs.msg import Recognition
-from image_recognition_openface.face_recognizer import FaceRecognizer
+from image_recognition_face_recognition.facenet_recognition import Facenet_recognition
 # Assign description to the help doc
+import math
 from sensor_msgs.msg import RegionOfInterest
 
 from image_recognition_util import image_writer
@@ -28,20 +29,26 @@ args = parser.parse_args()
 img = cv2.imread(args.image)
 
 # Create openface interface
-face_recognizer = FaceRecognizer(args.align_path, args.net_path)
+face_recognizer = Facenet_recognition()
 
 if args.db:
     face_recognizer.restore_trained_faces(args.db)
 
-recognized_faces = face_recognizer.recognize(img)
+recognized_faces = face_recognizer.face_detection(img)
 print(recognized_faces)
-annotated_original_image = image_writer.get_annotated_cv_image(img, [Recognition(
-    roi=RegionOfInterest(
-        x_offset=f.roi.x_offset,
-        y_offset=f.roi.y_offset,
-        width=f.roi.width,
-        height=f.roi.height
+
+recognitions = []
+for fr in recognized_faces:
+    face_recognition = [math.floor(xi) for xi in fr]
+    recognitions.append(Recognition(
+        roi=RegionOfInterest(
+            x_offset=face_recognition[0],
+            y_offset=face_recognition[1],
+            width=face_recognition[2] - face_recognition[0],
+            height=face_recognition[3] - face_recognition[1],
+        )
     )
-) for f in recognized_faces])
+)
+    annotated_original_image = image_writer.get_annotated_cv_image(img, recognitions) 
 cv2.imshow("result", annotated_original_image)
 cv2.waitKey()