Add emotion-recognition-demo

kubeedge-emotion-recognition-demo/README.md

@@ -0,0 +1,66 @@
+# KubeEdge Emotion Recognition Demo
+
+## Description
+
+KubeEdge Emotion Recognition is demo of emotion recognition with raspberry Pi. Raspberry Pi is equipped with an RS/E4 standard camera.
+
+<img src="images/raspberry-video.jpg">
+
+## Deployment Diagram 
+
+<img src="images/emotion-recognition-arch.jpg">
+
+## Prerequisites
+
+### Hardware Prerequisites
+
+1. RaspBerry-Pi (RaspBerry-Pi 4 has been used for this demo)
+2. RS/E4 Camera
+3. Display
+
+## Steps to reproduce
+
+1. Clone the kubeedge/examples repository.
+
+```console
+git clone https://github.com/kubeedge/examples.git /root/examples
+```
+
+2. Deploy Emotion Recognition Server.
+
+```console
+kubectl apply -f  /root/examples/kubeedge-emotion-recognition-demo/emotion-server.yaml
+```
+
+3. Deploy Emotion Recognition Client To raspberry Pi.
+
+```console
+kubectl apply -f  /root/examples/kubeedge-emotion-recognition-demo/emotion-client.yaml
+```
+```
+    spec:
+      containers:
+      - env:
+        - name: DISPLAY
+          value: :0
+        - name: FACEEMOTION_SERVER
+          value: $EMOTION_SERVER    #set emotion server address 
+        - name: FACEEMOTION_PORT
+          value: $EMOTION_CLIENT    #set emotion server port
+        volumeMounts:
+        - mountPath: /tmp/.X11-unix 
+          name: x11
+        - mountPath: /dev/video0
+          name: video
+        securityContext:
+          privileged: true
+      volumes:
+      - name: x11
+        hostPath:
+          path: /tmp/.X11-unix     #match your display device
+      - name: video
+        hostPath:
+          path: /dev/video0        #match your camera device
+```
+
+**Note:**  Only when a face is detected , the expression will appear on the display. 

kubeedge-emotion-recognition-demo/client/Dockerfile

@@ -0,0 +1,9 @@
+From kuramal/onnxruntime_opencv:arm32
+
+RUN pip install requests
+
+ADD ./emotion.py /home
+
+ADD ./haarcascade_frontalface_default.xml /home
+
+ENTRYPOINT [ "python", "/home/emotion.py" ]

kubeedge-emotion-recognition-demo/client/emotion.py

@@ -0,0 +1,72 @@
+#!/usr/bin/env python3
+# -*- coding: utf-8 -*-
+"""
+Created on Tue Jan  7 03:18:26 2020
+
+@author: gaowei
+"""
+
+import os
+import base64
+import json
+import requests
+import cv2
+import numpy as np
+
+faceCascade = cv2.CascadeClassifier('/home/haarcascade_frontalface_default.xml')
+
+env_dist = os.environ
+
+faceemotion_server = env_dist.get('FACEEMOTION_SERVER')
+faceemotion_port = env_dist.get('FACEEMOTION_PORT')
+
+cap = cv2.VideoCapture(0)
+timeF = 10
+count = 0
+request_url = "http://%s:%s/model/methods/predict" % (faceemotion_server, faceemotion_port)
+init = 0
+
+while(True):
+    count = count + 1
+    headers = {'accept': 'application/json','content-type': 'application/json'}
+    ret, frame = cap.read()
+    if (count%timeF != 0):
+        continue
+
+    img = cv2.flip(frame, 1)
+    gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
+    faces = faceCascade.detectMultiScale(
+        gray,     
+        scaleFactor=1.2,
+        minNeighbors=5,     
+        minSize=(20, 20)
+    )
+
+    if len(faces) == 0 and init == 1:
+        continue
+
+    init = 1
+
+    small_frame = cv2.resize(frame,(0,0),fx = 0.88,fy = 0.88)
+    image = cv2.imencode('.jpg',small_frame)[1]
+    img_BASE64 = str(base64.b64encode(image))[2:-1]
+
+    post_data = {"img_base64": img_BASE64}   
+    data = json.dumps(post_data).encode(encoding = 'utf-8')
+    response = requests.post(url=request_url, data=data, headers=headers)
+    res = json.loads(response.text)#dict
+    res_dict = json.loads(res["value"])
+    src = res_dict["img_url"]
+    data_processd = src.split(',')[1]
+    # base64解码
+    image_data = base64.b64decode(data_processd)
+    # 转换为np数组
+    img_array = np.fromstring(image_data, np.uint8)
+    # 转换成opencv可用格式
+    img = cv2.imdecode(img_array, cv2.COLOR_RGB2BGR)
+    cv2.imshow('img', img)
+    # Display the resulting frame
+    if cv2.waitKey(1) & 0xFF == ord('q'):
+        break
+cap.release()
+cv2.destroyAllWindows()