原文链接:应用 | OpenCV + OpenVINO实现人脸表情识别_英特尔边缘计算社区-CSDN博客
1.环境准备
先安装python 3.6.1,再在cmd里运行下面指令
python -m pip install --upgrade pip
pip install -i https://pypi.tuna.tsinghua.edu.cn/simple opencv-python
pip install -i https://pypi.tuna.tsinghua.edu.cn/simple openvino
python main.py
@2022.1.18添加说明
如果按照以上步奏,出现“from .ie_api import * ” 模块未找到的问题,可能是你的VS环境没装好,建议装2005到2019的VC运行环境
2.python代码
import cv2 as cv
import numpy as np
from openvino.inference_engine import IENetwork, IECore
weight_pb = "./opencv/opencv_face_detector_uint8.pb";
config_text = "./opencv/opencv_face_detector.pbtxt";
model_xml = "./openvino/emotions-recognition-retail-0003/emotions-recognition-retail-0003.xml"
model_bin = "./openvino/emotions-recognition-retail-0003/emotions-recognition-retail-0003.bin"
labels = ['neutral', 'happy', 'sad', 'surprise', 'anger']
emotion_labels = ["neutral", "anger", "disdain", "disgust", "fear", "happy", "sad", "surprise"]
emotion_net = IENetwork(model=model_xml, weights=model_bin)
ie = IECore()
versions = ie.get_versions("CPU")
input_blob = next(iter(emotion_net.inputs))
n, c, h, w = emotion_net.inputs[input_blob].shape
print(emotion_net.inputs[input_blob].shape)
output_info = emotion_net.outputs[next(iter(emotion_net.outputs.keys()))]
output_info.precision = "FP32"
exec_net = ie.load_network(network=emotion_net, device_name="CPU")
root_dir = "D:/facedb/emotion_dataset/"
count_ = 1
def emotion_detect(frame):
net = cv.dnn.readNetFromTensorflow(weight_pb, config=config_text)
h, w, c = frame.shape
blobImage = cv.dnn.blobFromImage(frame, 1.0, (300, 300), (104.0, 177.0, 123.0), False, False);
net.setInput(blobImage)
cvOut = net.forward()
# 绘制检测矩形
for detection in cvOut[0,0,:,:]:
score = float(detection[2])
if score > 0.5:
left = detection[3]*w
top = detection[4]*h
right = detection[5]*w
bottom = detection[6]*h
# roi and detect landmark
y1 = np.int32(top) if np.int32(top) > 0 else 0
y2 = np.int32(bottom) if np.int32(bottom) < h else h-1
x1 = np.int32(left) if np.int32(left) > 0 else 0
x2 = np.int32(right) if np.int32(right) < w else w-1
roi = frame[y1:y2,x1:x2,:]
image = cv.resize(roi, (64, 64))
image = image.transpose((2, 0, 1)) # Change data layout from HWC to CHW
res = exec_net.infer(inputs={input_blob: [image]})
prob_emotion = res['prob_emotion']
probs = np.reshape(prob_emotion, (5))
txt = labels[np.argmax(probs)]
cv.putText(frame, txt, (np.int32(left), np.int32(top)), cv.FONT_HERSHEY_SIMPLEX, 1.0, (255, 0, 0), 2)
cv.rectangle(frame, (np.int32(left), np.int32(top)),
(np.int32(right), np.int32(bottom)), (0, 0, 255), 2, 8, 0)
if __name__ == "__main__":
capture = cv.VideoCapture(0)
while True:
ret, frame = capture.read()
if ret is not True:
continue
emotion_detect(frame)
cv.imshow("emotion-detect-demo", frame)
c = cv.waitKey(30)
if c == 27:
break
工程链接:阿里云盘分享
说明:本工程使用PyCharm Community Edition 2021.2.3