coco关键点标注json_OpenCV图像处理基于OpenPose的关键点检测

import os
import cv2
import time
import numpy as np
import matplotlib.pyplot as pltclass general_pose_model(object):def __init__(self, modelpath):self.num_points &＃61; 22self.point_pairs &＃61; [[0,1],[1,2],[2,3],[3,4],[0,5],[5,6],[6,7],[7,8],[0,9],[9,10],[10,11],[11,12],[0,13],[13,14],[14,15],[15,16],[0,17],[17,18],[18,19],[19,20]]self.inHeight &＃61; 368self.threshold &＃61; 0.1self.hand_net &＃61; self.get_hand_model(modelpath)# 模型加载def get_hand_model(self, modelpath):prototxt &＃61; os.path.join(modelpath, "pose_deploy.prototxt")caffemodel &＃61; os.path.join(modelpath, "../pose_iter_102000.caffemodel")hand_model &＃61; cv2.dnn.readNetFromCaffe(prototxt, caffemodel)return hand_model# 预测def predict(self, imgfile):img_cv2 &＃61; cv2.imread(imgfile)img_height, img_width, _ &＃61; img_cv2.shapeaspect_ratio &＃61; img_width / img_heightinWidth &＃61; int(((aspect_ratio * self.inHeight) * 8) // 8)inpBlob &＃61; cv2.dnn.blobFromImage(img_cv2, 1.0 / 255, (inWidth, self.inHeight), (0, 0, 0), swapRB&＃61;False, crop&＃61;False)self.hand_net.setInput(inpBlob)output &＃61; self.hand_net.forward()# vis heatmapsself.vis_heatmaps(imgfile, output)points &＃61; []for idx in range(self.num_points):probMap &＃61; output[0, idx, :, :] # confidence map.probMap &＃61; cv2.resize(probMap, (img_width, img_height))# Find global maxima of the probMap.minVal, prob, minLoc, point &＃61; cv2.minMaxLoc(probMap)if prob > self.threshold:points.append((int(point[0]), int(point[1])))else:points.append(None)return points# heatmap可视化def vis_heatmaps(self, imgfile, net_outputs):img_cv2 &＃61; cv2.imread(imgfile)plt.figure(figsize&＃61;[10, 10])for pdx in range(self.num_points):probMap &＃61; net_outputs[0, pdx, :, :]probMap &＃61; cv2.resize(probMap, (img_cv2.shape[1], img_cv2.shape[0]))plt.subplot(5, 5, pdx&＃43;1)plt.imshow(cv2.cvtColor(img_cv2, cv2.COLOR_BGR2RGB))plt.imshow(probMap, alpha&＃61;0.6)plt.colorbar()plt.axis("off")plt.show()# 手部关键点可视化def vis_pose(self, imgfile, points):img_cv2 &＃61; cv2.imread(imgfile)img_cv2_copy &＃61; np.copy(img_cv2)for idx in range(len(points)):if points[idx]:cv2.circle(img_cv2_copy, points[idx], 8, (0, 255, 255), thickness&＃61;-1,lineType&＃61;cv2.FILLED)cv2.putText(img_cv2_copy, "{}".format(idx), points[idx], cv2.FONT_HERSHEY_SIMPLEX,1, (0, 0, 0), 2, lineType&＃61;cv2.LINE_AA)# 绘制连接点for pair in self.point_pairs:partA &＃61; pair[0]partB &＃61; pair[1]if points[partA] and points[partB]:cv2.line(img_cv2, points[partA], points[partB], (0, 255, 255), 3)cv2.circle(img_cv2, points[partA], 8, (0, 0, 255), thickness&＃61;-1, lineType&＃61;cv2.FILLED)plt.figure(figsize&＃61;[10, 10])plt.subplot(1, 2, 1)plt.imshow(cv2.cvtColor(img_cv2, cv2.COLOR_BGR2RGB))plt.axis("off")plt.subplot(1, 2, 2)plt.imshow(cv2.cvtColor(img_cv2_copy, cv2.COLOR_BGR2RGB))plt.axis("off")plt.show()if __name__ &＃61;&＃61; &＃39;__main__&＃39;:print("[INFO]Pose estimation.")os.listdir(imgs_path)]img_files &＃61; [&＃39;hand.jpg&＃39;]start &＃61; time.time()modelpath &＃61; ""pose_model &＃61; general_pose_model(modelpath)print("[INFO]Model loads time: ", time.time() - start)for img_file in img_files:start &＃61; time.time()res_points &＃61; pose_model.predict(img_file)print("[INFO]Model predicts time: ", time.time() - start)pose_model.vis_pose(img_file, res_points)print("[INFO]Done.")

BODY25: http://posefs1.perception.cs.cmu.edu/OpenPose/models/pose/body_25/pose_iter_584000.caffemodel
COCO: http://posefs1.perception.cs.cmu.edu/OpenPose/models/pose/coco/pose_iter_440000.caffemodel
MPI: http://posefs1.perception.cs.cmu.edu/OpenPose/models/pose/mpi/pose_iter_160000.caffemodelCOCO prototxt&＃xff1a;https://github.com/CMU-Perceptual-Computing-Lab/openpose/blob/master/models/pose/coco/pose_deploy_linevec.prototxt

import cv2
import time
import numpy as np
import matplotlib.pyplot as plt
import osclass general_pose_model(object):def __init__(self, modelpath, mode&＃61;"BODY25"):# 指定采用的模型# Body25: 25 points# COCO: 18 points# MPI: 15 pointsself.inWidth &＃61; 368self.inHeight &＃61; 368self.threshold &＃61; 0.1if mode &＃61;&＃61; "BODY25":self.pose_net &＃61; self.general_body25_model(modelpath)elif mode &＃61;&＃61; "COCO":self.pose_net &＃61; self.general_coco_model(modelpath)elif mode &＃61;&＃61; "MPI":self.pose_net &＃61; self.get_mpi_model(modelpath)def get_mpi_model(self, modelpath):self.points_name &＃61; { "Head": 0, "Neck": 1, "RShoulder": 2, "RElbow": 3, "RWrist": 4,"LShoulder": 5, "LElbow": 6, "LWrist": 7, "RHip": 8, "RKnee": 9, "RAnkle": 10, "LHip": 11, "LKnee": 12, "LAnkle": 13, "Chest": 14, "Background": 15 }self.num_points &＃61; 15self.point_pairs &＃61; [[0, 1], [1, 2], [2, 3], [3, 4], [1, 5], [5, 6], [6, 7], [1, 14],[14, 8], [8, 9], [9, 10], [14, 11], [11, 12], [12, 13]]prototxt &＃61; os.path.join(modelpath,"pose/mpi/pose_deploy_linevec_faster_4_stages.prototxt")caffemodel &＃61; os.path.join(modelpath, "pose/mpi/pose_iter_160000.caffemodel")mpi_model &＃61; cv2.dnn.readNetFromCaffe(prototxt, caffemodel)return mpi_modeldef general_coco_model(self, modelpath):self.points_name &＃61; {"Nose": 0, "Neck": 1, "RShoulder": 2, "RElbow": 3, "RWrist": 4,"LShoulder": 5, "LElbow": 6, "LWrist": 7, "RHip": 8, "RKnee": 9, "RAnkle": 10, "LHip": 11, "LKnee": 12, "LAnkle": 13, "REye": 14, "LEye": 15, "REar": 16, "LEar": 17, "Background": 18}self.num_points &＃61; 18self.point_pairs &＃61; [[1, 0], [1, 2], [1, 5], [2, 3], [3, 4], [5, 6], [6, 7], [1, 8], [8, 9],[9, 10], [1, 11], [11, 12], [12, 13], [0, 14], [0, 15], [14, 16], [15, 17]]prototxt &＃61; os.path.join(modelpath, "openpose_pose_coco.prototxt")caffemodel &＃61; os.path.join(modelpath, "../pose_iter_440000.caffemodel")print(prototxt, caffemodel)coco_model &＃61; cv2.dnn.readNetFromCaffe(prototxt, caffemodel)return coco_modeldef general_body25_model(self, modelpath):self.num_points &＃61; 25self.point_pairs &＃61; [[1, 0], [1, 2], [1, 5], [2, 3], [3, 4], [5, 6], [6, 7], [0, 15], [15, 17], [0, 16], [16, 18], [1, 8],[8, 9], [9, 10], [10, 11], [11, 22], [22, 23], [11, 24],[8, 12], [12, 13], [13, 14], [14, 19], [19, 20], [14, 21]]prototxt &＃61; os.path.join(modelpath, "pose/body_25/pose_deploy.prototxt")caffemodel &＃61; os.path.join(modelpath, "pose/body_25/pose_iter_584000.caffemodel")coco_model &＃61; cv2.dnn.readNetFromCaffe(prototxt, caffemodel)return coco_modeldef predict(self, imgfile):img_cv2 &＃61; cv2.imread(imgfile)img_height, img_width, _ &＃61; img_cv2.shapeinpBlob &＃61; cv2.dnn.blobFromImage(img_cv2, 1.0 / 255, (self.inWidth, self.inHeight),(0, 0, 0), swapRB&＃61;False, crop&＃61;False)self.pose_net.setInput(inpBlob)self.pose_net.setPreferableBackend(cv2.dnn.DNN_BACKEND_OPENCV)self.pose_net.setPreferableTarget(cv2.dnn.DNN_TARGET_OPENCL)output &＃61; self.pose_net.forward()H &＃61; output.shape[2]W &＃61; output.shape[3]print(output.shape)# vis heatmapsself.vis_heatmaps(img_file, output)#points &＃61; []for idx in range(self.num_points):probMap &＃61; output[0, idx, :, :] # confidence map.# Find global maxima of the probMap.minVal, prob, minLoc, point &＃61; cv2.minMaxLoc(probMap)# Scale the point to fit on the original imagex &＃61; (img_width * point[0]) / Wy &＃61; (img_height * point[1]) / Hif prob > self.threshold:points.append((int(x), int(y)))else:points.append(None)return pointsdef vis_heatmaps(self, imgfile, net_outputs):img_cv2 &＃61; cv2.imread(imgfile)plt.figure(figsize&＃61;[10, 10])for pdx in range(self.num_points):probMap &＃61; net_outputs[0, pdx, :, :]probMap &＃61; cv2.resize(probMap, (img_cv2.shape[1], img_cv2.shape[0]))plt.subplot(5, 5, pdx&＃43;1)plt.imshow(cv2.cvtColor(img_cv2, cv2.COLOR_BGR2RGB))plt.imshow(probMap, alpha&＃61;0.6)plt.colorbar()plt.axis("off")plt.show()def vis_pose(self, imgfile, points):img_cv2 &＃61; cv2.imread(imgfile)img_cv2_copy &＃61; np.copy(img_cv2)for idx in range(len(points)):if points[idx]:cv2.circle(img_cv2_copy, points[idx], 3, (0, 0, 255), thickness&＃61;-1,lineType&＃61;cv2.FILLED)cv2.putText(img_cv2_copy, "{}".format(idx), points[idx], cv2.FONT_HERSHEY_SIMPLEX,.6, (0, 255, 255), 1, lineType&＃61;cv2.LINE_AA)# Draw Skeletonfor pair in self.point_pairs:partA &＃61; pair[0]partB &＃61; pair[1]if points[partA] and points[partB]:cv2.line(img_cv2, points[partA], points[partB], (0, 255, 0), 3)cv2.circle(img_cv2, points[partA], 3, (0, 0, 255), thickness&＃61;-1, lineType&＃61;cv2.FILLED)