星期

2020年09月08日

人脸识别+表情检测+行人检测+人脸关键点检测!Open CV

2020-09-08 15:42:28 来源:互联网 阅读:-

一 环境搭配

环境安装请看上一篇博客:传送门

二 人脸识别与关键点检测

以下来源于Openvino官方model ,在win10 和ubuntu大体步骤相似,跑demo:
想转ubuntu 或者win10 方法一样,我下面分别用win10 和ubuntu跑几个demo,大家可以试着做一下。
效果展示

人脸识别+表情检测+行人检测+人脸关键点检测!Open CV

一、 准备流程:

  1. 在python环境中加载openvino

打开openvino安装目录如:
C:\Intel\openvino\python\python3.6

把目录下的openvino文件夹复制到

系统的python环境安装目录下如: C:\Python36\Lib\site-packages
2. 编译

C:\Intel\openvino\deployment_tools\inference_engine\samples 路径下执行:

build_samples_msvc2017.bat

执行完后在

C:\Users\kang\Documents\Intel\OpenVINO 目录

可以看到生成的

inference_engine_samples_build_2017 文件目录

在build目录中也可以找到cpu_extension:

cpu_extension = “C:\Users\kang\Documents\Intel\OpenVINO\inference_engine_samples_build_2017\intel64\Release\cpu_extension.dll”

  1. 下载模型,记录路径

face-detection-adas-0001

landmarks-regression-retail-0009

记录xml地址

model_xml = “”
model_bin = “”

二、 参数说明

  1. 人脸检测基于MobileNet v1版本
    输入格式:[1x3x384x672] = BCHW输出格式:[1,1,N,7] = [image_id, label, conf, x_min, y_min, x_max, y_max]
  2. landmark提取landmark提取 - 基于卷积神经网络,提取5个点
    输入 [1x3x48x48] = BCHW输出 [1X10X1X1] = 五个点坐标(x0,y0,x1,y1…x4,y4)
  3. python版本的api介绍同步调用,执行输入
    Im_exec_net.infer(inputs={“0”:face_roi})
  4. 获取输出landmark_res = Im_exec_net.request[0].outputs[Im_output_blob]
    landmark_res = np.reshape(landmark_res,(5,2))

三、 附录代码:

import sys
import cv2import numpy as npimport timeimport logging as logfrom openvino.inference_engine import IENetwork, IEPluginmodel_xml = "C:/Users/kang/Downloads/open_model_zoo-2019/model_downloader/Transportation/object_detection/face/pruned_mobilenet_reduced_ssd_shared_weights/dldt/face-detection-adas-0001.xml"
model_bin = "C:/Users/kang/Downloads/open_model_zoo-2019/model_downloader/Transportation/object_detection/face/pruned_mobilenet_reduced_ssd_shared_weights/dldt/face-detection-adas-0001.bin"
plugin_dir = "C:/Intel/openvino/deployment_tools/inference_engine/bin/intel64/Release"
cpu_extension = "C:/Users/kang/Documents/Intel/OpenVINO/inference_engine_samples_build_2017/intel64/Release/cpu_extension.dll"
landmark_xml = "C:/Users/kang/Downloads/open_model_zoo-2019/model_downloader/Retail/object_attributes/landmarks_regression/0009/dldt/landmarks-regression-retail-0009.xml"
landmark_bin = "C:/Users/kang/Downloads/open_model_zoo-2019/model_downloader/Retail/object_attributes/landmarks_regression/0009/dldt/landmarks-regression-retail-0009.bin"
def face_landmark_demo():    log.basicConfig(format="[ %(levelname)s ] %(message)s",
                   level=log.INFO,
                   stream=sys.stdout)
   # Plugin initialization for specified device and load extensions library if specified
   log.info("Initializing plugin for {} device...".format("CPU"))
   plugin = IEPlugin(device="CPU", plugin_dirs=plugin_dir)
   plugin.add_cpu_extension(cpu_extension)    # lut    lut = []    lut.append((0, 0, 255))
   lut.append((255, 0, 0))
   lut.append((0, 255, 0))
   lut.append((0, 255, 255))
   lut.append((255, 0, 255))
   # Read IR    log.info("Reading IR...")
   net = IENetwork(model=model_xml, weights=model_bin)    landmark_net = IENetwork(model=landmark_xml, weights=landmark_bin)    if plugin.device == "CPU":
       supported_layers = plugin.get_supported_layers(net)        not_supported_layers = [            l for l in net.layers.keys() if l not in supported_layers
       ]        if len(not_supported_layers) != 0:
           log.error(
               "Following layers are not supported by the plugin for specified device {}:\n {}"
               .format(plugin.device, ', '.join(not_supported_layers)))
           log.error(
               "Please try to specify cpu extensions library path in demo's command line parameters using -l "
               "or --cpu_extension command line argument")
           sys.exit(1)
   assert len(
       net.inputs.keys()) == 1, "Demo supports only single input topologies"
   assert len(net.outputs) == 1, "Demo supports only single output topologies"
   input_blob = next(iter(net.inputs))
   out_blob = next(iter(net.outputs))
   lm_input_blob = next(iter(landmark_net.inputs))
   lm_out_blob = next(iter(landmark_net.outputs))
   log.info("Loading IR to the plugin...")
   exec_net = plugin.load(network=net, num_requests=2)
   lm_exec_net = plugin.load(network=landmark_net)
   # Read and pre-process input image
   n, c, h, w = net.inputs[input_blob].shape    nm, cm, hm, wm = landmark_net.inputs[lm_input_blob].shape    del net    del landmark_net    cap = cv2.VideoCapture("C:/Users/kang/Downloads/material/av77002671.mp4")
   cur_request_id = 0
   next_request_id = 1
   log.info("Starting inference in async mode...")
   log.info("To switch between sync and async modes press Tab button")
   log.info("To stop the demo execution press Esc button")
   is_async_mode = True    render_time = 0
   ret, frame = cap.read()
   print(
       "To close the application, press 'CTRL+C' or any key with focus on the output window"
   )    while cap.isOpened():
       if is_async_mode:
           ret, next_frame = cap.read()
       else:
           ret, frame = cap.read()
       if not ret:
           break
       initial_w = cap.get(3)
       initial_h = cap.get(4)
       inf_start = time.time()
       if is_async_mode:
           in_frame = cv2.resize(next_frame, (w, h))            in_frame = in_frame.transpose(                (2, 0, 1))  # Change data layout from HWC to CHW
           in_frame = in_frame.reshape((n, c, h, w))            exec_net.start_async(request_id=next_request_id,                                inputs={input_blob: in_frame})        else:
           in_frame = cv2.resize(frame, (w, h))            in_frame = in_frame.transpose(                (2, 0, 1))  # Change data layout from HWC to CHW
           in_frame = in_frame.reshape((n, c, h, w))            exec_net.start_async(request_id=cur_request_id,                                inputs={input_blob: in_frame})        if exec_net.requests[cur_request_id].wait(-1) == 0:
           res = exec_net.requests[cur_request_id].outputs[out_blob]            for obj in res[0][0]:
               if obj[2] > 0.5:
                   xmin = int(obj[3] * initial_w)
                   ymin = int(obj[4] * initial_h)
                   xmax = int(obj[5] * initial_w)
                   ymax = int(obj[6] * initial_h)
                   if xmin > 0 and ymin > 0 and (xmax < initial_w) and (
                           ymax < initial_h):                        roi = frame[ymin:ymax, xmin:xmax, :]                        rh, rw = roi.shape[:2]
                       face_roi = cv2.resize(roi, (wm, hm))                        face_roi = face_roi.transpose((2, 0, 1))
                       face_roi = face_roi.reshape((nm, cm, hm, wm))                        lm_exec_net.infer(inputs={'0': face_roi})
                       landmark_res = lm_exec_net.requests[0].outputs[
                           lm_out_blob]                        landmark_res = np.reshape(landmark_res, (5, 2))
                       for m in range(len(landmark_res)):
                           x = landmark_res[m][0] * rw
                           y = landmark_res[m][1] * rh
                           cv2.circle(roi, (np.int32(x), np.int32(y)), 3,
                                   lut[m], 2, 8, 0)
                   cv2.rectangle(frame, (xmin, ymin), (xmax, ymax),                                (0, 0, 255), 2, 8, 0)
           inf_end = time.time()
           det_time = inf_end - inf_start            # Draw performance stats            inf_time_message = "Inference time: {:.3f} ms, FPS:{:.3f}".format(
               det_time * 1000, 1000 / (det_time * 1000 + 1))
           render_time_message = "OpenCV rendering time: {:.3f} ms".format(
               render_time * 1000)
           async_mode_message = "Async mode is on. Processing request {}".format(cur_request_id) if is_async_mode else \
               "Async mode is off. Processing request {}".format(cur_request_id)
           cv2.putText(frame, inf_time_message, (15, 15),
                       cv2.FONT_HERSHEY_COMPLEX, 0.5, (200, 10, 10), 1)
           cv2.putText(frame, render_time_message, (15, 30),
                       cv2.FONT_HERSHEY_COMPLEX, 0.5, (10, 10, 200), 1)
           cv2.putText(frame, async_mode_message, (10, int(initial_h - 20)),
                       cv2.FONT_HERSHEY_COMPLEX, 0.5, (10, 10, 200), 1)
       render_start = time.time()
       cv2.imshow("face detection", frame)
       render_end = time.time()
       render_time = render_end - render_start        if is_async_mode:
           cur_request_id, next_request_id = next_request_id, cur_request_id            frame = next_frame        key = cv2.waitKey(1)
       if key == 27:
           break
   cv2.destroyAllWindows()    del exec_net    del lm_exec_net    del pluginif __name__ == '__main__':
   sys.exit(face_landmark_demo() or 0)

三 行人检测

1.测试环境:

ubuntu版本:18.04.1LTS

openvino版本:2020.1.023

模型文档链接:https://docs.openvinotoolkit.org/latest/_demos_crossroad_camera_demo_README.html
2.下载模型

进入open_model_zoo路径

cd /home/kang/open_model_zoo/tools/downloader

在模型列表中找到要下载的模型并下载:

./downloader.py --name person-vehicle-bike-detection-crossroad-0078

记录xml文件下载路径:

/home/kang/open_model_zoo/tools/downloader/intel/person-vehicle-bike-detection-crossroad-0078/FP32/person-vehicle-bike-detection-crossroad-0078.xml
3.编译

执行下列命令

cd /opt/intel/openvino/deployment_tools/inference_engine/demos
./build_demos.sh123
人脸识别+表情检测+行人检测+人脸关键点检测!Open CV

进入crossroad_camera_demo路径,执行make

cd  ~/omz_demos_build/crossroad_camera_demo
make -j412
人脸识别+表情检测+行人检测+人脸关键点检测!Open CV

3.运行

cd ~/omz_demos_build/intel64/Release
./crossroad_camera_demo -m /home/kang/open_model_zoo/tools/downloader/intel/person-vehicle-bike-detection-crossroad-0078/FP32/person-vehicle-bike-detection-crossroad-0078.xml -d CPU -i /home/kang/Downloads/test_data/pedestrian.png
123

得到结果和图像信息。

人脸识别+表情检测+行人检测+人脸关键点检测!Open CV


同样也可以将xml进行python运行。

四 人脸识别与表情检测

效果展示

人脸识别+表情检测+行人检测+人脸关键点检测!Open CV

一、 准备流程:

  1. 在python环境中加载openvino

打开openvino安装目录如:
C:\Intel\openvino\python\python3.6

把目录下的openvino文件夹复制到

系统的python环境安装目录下如: C:\Python36\Lib\site-packages
2. 编译

C:\Intel\openvino\deployment_tools\inference_engine\samples 路径下执行:

build_samples_msvc2017.bat

执行完后在

C:\Users\kang\Documents\Intel\OpenVINO 目录

可以看到生成的

inference_engine_samples_build_2017 文件目录

在build目录中也可以找到cpu_extension:

cpu_extension = “C:\Users\kang\Documents\Intel\OpenVINO\inference_engine_samples_build_2017\intel64\Release\cpu_extension.dll”

  1. 下载模型,记录路径

face-detection-adas-0001

emotions-recognition-retail-0003

model_xml = “”
model_bin = “”

二、 参数介绍:

  1. emotions提取基于MobileNet v1版本
    · 输入格式:[1x3x384x672]= BCHW· 输出格式:[1, 1, N, 7] = [image_id, label, conf, x_min, y_min, x_max, y_max]表情识别网络 – 输入-[1x3x64x64]=BCHW· 输出格式- [1, 5, 1, 1]· 检测五种表情 (‘neutral’, ‘happy’, ‘sad’, ‘surprise’, ‘anger’)
  2. python版本的api介绍同步调用,执行输入
    landmark_res = exec_emotions_net.infer(inputs={input_blob: [face_roi]})
  3. 获取输出landmark_res = landmark_res[‘prob_emotion’]
    landmark_res = np.reshape(landmark_res, (5))landmark_res = labels[np.argmax(landmark_res)]

代码:

import sys
import cv2import numpy as npimport timeimport logging as logfrom openvino.inference_engine import IENetwork, IEPluginplugin_dir = "C:/Intel/openvino/deployment_tools/inference_engine/bin/intel64/Release"
cpu_extension = "C:/Users/kang/Documents/Intel/OpenVINO/inference_engine_samples_build_2017/intel64/Release/cpu_extension.dll"
# face-detection-adas-0001
model_xml  = "C:/Users/kang/Downloads/openvino_sample_show/open_model_zoo/model_downloader/Transportation/object_detection/face/pruned_mobilenet_reduced_ssd_shared_weights/dldt/face-detection-adas-0001.xml"
model_bin = "C:/Users/kang/Downloads/openvino_sample_show/open_model_zoo/model_downloader/Transportation/object_detection/face/pruned_mobilenet_reduced_ssd_shared_weights/dldt/face-detection-adas-0001.bin"
# emotions-recognition-retail-0003
emotions_xml = "C:/Users/kang/Downloads/openvino_sample_show/open_model_zoo/model_downloader/Retail/object_attributes/emotions_recognition/0003/dldt/emotions-recognition-retail-0003.xml"
emotions_bin = "C:/Users/kang/Downloads/openvino_sample_show/open_model_zoo/model_downloader/Retail/object_attributes/emotions_recognition/0003/dldt/emotions-recognition-retail-0003.bin"
labels = ['neutral', 'happy', 'sad', 'surprise', 'anger']
def face_emotions_demo():    log.basicConfig(format="[ %(levelname)s ] %(message)s",
                   level=log.INFO,
                   stream=sys.stdout)
   # Plugin initialization for specified device and load extensions library if specified
   log.info("Initializing plugin for {} device...".format("CPU"))
   plugin = IEPlugin(device="CPU", plugin_dirs=plugin_dir)
   plugin.add_cpu_extension(cpu_extension)    # Read IR    log.info("Reading IR...")
   net = IENetwork(model=model_xml, weights=model_bin)    emotions_net = IENetwork(model=emotions_xml, weights=emotions_bin)    if plugin.device == "CPU":
       supported_layers = plugin.get_supported_layers(net)        not_supported_layers = [            l for l in net.layers.keys() if l not in supported_layers
       ]        if len(not_supported_layers) != 0:
           log.error(
               "Following layers are not supported by the plugin for specified device {}:\n {}"
               .format(plugin.device, ', '.join(not_supported_layers)))
           log.error(
               "Please try to specify cpu extensions library path in demo's command line parameters using -l "
               "or --cpu_extension command line argument")
           sys.exit(1)
   assert len(
       net.inputs.keys()) == 1, "Demo supports only single input topologies"
   assert len(net.outputs) == 1, "Demo supports only single output topologies"
   input_blob = next(iter(net.inputs))
   out_blob = next(iter(net.outputs))
   em_input_blob = next(iter(emotions_net.inputs))
   em_out_blob = next(iter(emotions_net.outputs))
   log.info("Loading IR to the plugin...")
   # 生成可执行网络,异步执行 num_requests=2
   exec_net = plugin.load(network=net, num_requests=2)
   exec_emotions_net = plugin.load(network=emotions_net)
   # Read and pre-process input image
   n, c, h, w = net.inputs[input_blob].shape    en, ec, eh, ew = emotions_net.inputs[em_input_blob].shape    del net    del emotions_net    cap = cv2.VideoCapture("C:/Users/kang/Downloads/openvino_sample_show/material/face_detection_demo.mp4")
   cur_request_id = 0
   next_request_id = 1
   log.info("Starting inference in async mode...")
   log.info("To switch between sync and async modes press Tab button")
   log.info("To stop the demo execution press Esc button")
   is_async_mode = True    render_time = 0
   ret, frame = cap.read()
   print(
       "To close the application, press 'CTRL+C' or any key with focus on the output window"
   )    while cap.isOpened():
       if is_async_mode:
           ret, next_frame = cap.read()
       else:
           ret, frame = cap.read()
       if not ret:
           break
       initial_w = cap.get(3)
       initial_h = cap.get(4)
       inf_start = time.time()
       if is_async_mode:
           in_frame = cv2.resize(next_frame, (w, h))            in_frame = in_frame.transpose(                (2, 0, 1))  # Change data layout from HWC to CHW
           in_frame = in_frame.reshape((n, c, h, w))            exec_net.start_async(request_id=next_request_id,                                inputs={input_blob: in_frame})        else:
           in_frame = cv2.resize(frame, (w, h))            in_frame = in_frame.transpose(                (2, 0, 1))  # Change data layout from HWC to CHW
           in_frame = in_frame.reshape((n, c, h, w))            exec_net.start_async(request_id=cur_request_id,                                inputs={input_blob: in_frame})        if exec_net.requests[cur_request_id].wait(-1) == 0:
           res = exec_net.requests[cur_request_id].outputs[out_blob]            # 输出格式:[1,1,N,7]  从N行人脸中找到7个值   = [image_id,label,conf,x_min,y_min,x_max,y_max]
           for obj in res[0][0]:
               if obj[2] > 0.5:
                   xmin = int(obj[3] * initial_w)
                   ymin = int(obj[4] * initial_h)
                   xmax = int(obj[5] * initial_w)
                   ymax = int(obj[6] * initial_h)
                   if xmin > 0 and ymin > 0 and (xmax < initial_w) and (ymax < initial_h):
                       roi = frame[ymin:ymax,xmin:xmax,:]                        face_roi = cv2.resize(roi,(ew,eh))                        face_roi =face_roi.transpose((2, 0, 1))
                       face_roi= face_roi.reshape((en, ec, eh, ew))                        # 解析结果                        landmark_res = exec_emotions_net.infer(inputs={input_blob: [face_roi]})                        landmark_res = landmark_res['prob_emotion']
                       landmark_res = np.reshape(landmark_res, (5))
                       landmark_res = labels[np.argmax(landmark_res)]                        cv2.putText(frame, landmark_res, (np.int32(xmin), np.int32(ymin)), cv2.FONT_HERSHEY_SIMPLEX, 1.0,
                               (255, 0, 0), 2)
                       cv2.rectangle(frame, (np.int32(xmin), np.int32(ymin)), (np.int32(xmax), np.int32(ymax)),                                    (0, 0, 255), 2, 8, 0)
                   cv2.rectangle(frame, (xmin,ymin), (xmax,ymax), (0, 0, 255), 2, 8, 0)
           inf_end = time.time()
           det_time = inf_end - inf_start            # Draw performance stats            inf_time_message = "Inference time: {:.3f} ms, FPS:{:.3f}".format(det_time * 1000, 1000 / (det_time*1000 + 1))
           render_time_message = "OpenCV rendering time: {:.3f} ms".format(
               render_time * 1000)
           async_mode_message = "Async mode is on. Processing request {}".format(cur_request_id) if is_async_mode else \
               "Async mode is off. Processing request {}".format(cur_request_id)
           cv2.putText(frame, inf_time_message, (15, 15),
                       cv2.FONT_HERSHEY_COMPLEX, 0.5, (200, 10, 10), 1)
           cv2.putText(frame, render_time_message, (15, 30),
                       cv2.FONT_HERSHEY_COMPLEX, 0.5, (10, 10, 200), 1)
           cv2.putText(frame, async_mode_message, (10, int(initial_h - 20)),
                       cv2.FONT_HERSHEY_COMPLEX, 0.5, (10, 10, 200), 1)
       render_start = time.time()
       cv2.imshow("face emotions demo", frame)
       render_end = time.time()
       render_time = render_end - render_start        if is_async_mode:
           cur_request_id, next_request_id = next_request_id, cur_request_id            frame = next_frame        key = cv2.waitKey(1)
       if key == 27:
           break
   cv2.destroyAllWindows()    del exec_net    del exec_emotions_net    del pluginif __name__ == '__main__':
   sys.exit(face_emotions_demo() or 0)


看不懂就对了,想获取更多视频教程源码私信小编01


推荐阅读:北方生活网