轻拔琴弦|人脸识别+表情检测+行人检测+人脸关键点检测！Open CV

环境安装请看上一篇博客：传送门
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

本文插图

一、准备流程：

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-packages2. 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”

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

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]
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)
python版本的api介绍同步调用，执行输入Im_exec_net.infer(inputs={“0”:face_roi})
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 sysimport 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 specifiedlog.info("Initializing plugin for {} device...".format("CPU"))plugin = IEPlugin(device="CPU", plugin_dirs=plugin_dir)plugin.add_cpu_extension(cpu_extension)# lutlut = []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 IRlog.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 imagen, c, h, w = net.inputs[input_blob].shapenm, cm, hm, wm = landmark_net.inputs[lm_input_blob].shapedel netdel landmark_netcap = cv2.VideoCapture("C:/Users/kang/Downloads/material/av77002671.mp4")cur_request_id = 0next_request_id = 1log.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 = Truerender_time = 0ret, 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:breakinitial_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 CHWin_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 CHWin_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] * rwy = landmark_res[m][1] * rhcv2.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 statsinf_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_startif is_async_mode:cur_request_id, next_request_id = next_request_id, cur_request_idframe = next_framekey = cv2.waitKey(1)if key == 27:breakcv2.destroyAllWindows()del exec_netdel lm_exec_netdel pluginif __name__ == '__main__':sys.exit(face_landmark_demo() or 0)

1.测试环境：
ubuntu版本：18.04.1LTS
openvino版本：2020.1.023
模型文档链接：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.xml3.编译
执行下列命令
cd /opt/intel/openvino/deployment_tools/inference_engine/demos ./build_demos.sh123

本文插图

进入crossroad_camera_demo路径，执行make
cd~/omz_demos_build/crossroad_camera_demomake -j412

本文插图

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.png123

得到结果和图像信息。

本文插图

同样也可以将xml进行python运行。
效果展示

本文插图

一、准备流程：

在python环境中加载openvino

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：

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’)
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]})
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 sysimport 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-0001model_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-0003emotions_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 specifiedlog.info("Initializing plugin for {} device...".format("CPU"))plugin = IEPlugin(device="CPU", plugin_dirs=plugin_dir)plugin.add_cpu_extension(cpu_extension)# Read IRlog.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=2exec_net = plugin.load(network=net, num_requests=2)exec_emotions_net = plugin.load(network=emotions_net)# Read and pre-process input imagen, c, h, w = net.inputs[input_blob].shapeen, ec, eh, ew = emotions_net.inputs[em_input_blob].shapedel netdel emotions_netcap = cv2.VideoCapture("C:/Users/kang/Downloads/openvino_sample_show/material/face_detection_demo.mp4")cur_request_id = 0next_request_id = 1log.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 = Truerender_time = 0ret, 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:breakinitial_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 CHWin_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 CHWin_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 statsinf_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_startif is_async_mode:cur_request_id, next_request_id = next_request_id, cur_request_idframe = next_framekey = cv2.waitKey(1)if key == 27:breakcv2.destroyAllWindows()del exec_netdel exec_emotions_netdel 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
【轻拔琴弦|人脸识别+表情检测+行人检测+人脸关键点检测！Open CV】