from pathlib import Path import matplotlib.path as mat from ultralytics.models.utils import ops from utils.general import strtolstl from utils.general import compute_IOU from torchvision import transforms from PIL import Image import torch import torch.nn.functional as F import numpy as np import torch.nn as nn from ultralytics.engine.results import Results from shapely.geometry import Point from shapely.geometry.polygon import Polygon from pydantic import BaseModel mean, std = [0.485, 0.456, 0.406], [0.229, 0.224, 0.225] test = transforms.Compose([transforms.Resize((224,224)), #transforms.CenterCrop(224), transforms.ToTensor(), transforms.Normalize(mean=mean, std=std) ]) def clapre(modelcla,claimg,clapoint): imgten = torch.stack(claimg,dim=0) clapoint = torch.stack(clapoint,dim=0) imgten = imgten.to(0) result = modelcla(imgten) result = F.softmax(result) print(result) index = result.argmax(1) index = index.cpu().numpy() index = np.argwhere(index<5) index = index.reshape(-1) print(index) if len(index)>0: print(clapoint[index]) return clapoint[index] else: return None class Model(nn.Module): def __init__(self, A, nnode, nfeature, nclass): super().__init__() self.fc1 = nn.Linear(nnode * nfeature, 512) self.fc2 = nn.Linear(512, nclass) def forward(self, x): x = x.view(-1, int(x.size(1) * x.size(2))) x = F.relu(self.fc1(x)) x = F.dropout(x, 0.7, training=self.training) return self.fc2(x) class GetKeypoint(BaseModel): NOSE: int = 0 LEFT_EYE: int = 1 RIGHT_EYE: int = 2 LEFT_EAR: int = 3 RIGHT_EAR: int = 4 LEFT_SHOULDER: int = 5 RIGHT_SHOULDER: int = 6 LEFT_ELBOW: int = 7 RIGHT_ELBOW: int = 8 LEFT_WRIST: int = 9 RIGHT_WRIST: int = 10 LEFT_HIP: int = 11 RIGHT_HIP: int = 12 LEFT_KNEE: int = 13 RIGHT_KNEE: int = 14 LEFT_ANKLE: int = 15 RIGHT_ANKLE: int = 16 def extract_keypoint(get_keypoint, keypoint): # nose nose_x, nose_y = keypoint[get_keypoint.NOSE] # eye # left_eye_x, left_eye_y = keypoint[get_keypoint.LEFT_EYE] # right_eye_x, right_eye_y = keypoint[get_keypoint.RIGHT_EYE] # # ear # left_ear_x, left_ear_y = keypoint[get_keypoint.LEFT_EAR] # right_ear_x, right_ear_y = keypoint[get_keypoint.RIGHT_EAR] # shoulder left_shoulder_x, left_shoulder_y = keypoint[get_keypoint.LEFT_SHOULDER] right_shoulder_x, right_shoulder_y = keypoint[get_keypoint.RIGHT_SHOULDER] # elbow left_elbow_x, left_elbow_y = keypoint[get_keypoint.LEFT_ELBOW] right_elbow_x, right_elbow_y = keypoint[get_keypoint.RIGHT_ELBOW] # wrist left_wrist_x, left_wrist_y = keypoint[get_keypoint.LEFT_WRIST] right_wrist_x, right_wrist_y = keypoint[get_keypoint.RIGHT_WRIST] # hip left_hip_x, left_hip_y = keypoint[get_keypoint.LEFT_HIP] right_hip_x, right_hip_y = keypoint[get_keypoint.RIGHT_HIP] # knee left_knee_x, left_knee_y = keypoint[get_keypoint.LEFT_KNEE] right_knee_x, right_knee_y = keypoint[get_keypoint.RIGHT_KNEE] # ankle left_ankle_x, left_ankle_y = keypoint[get_keypoint.LEFT_ANKLE] right_ankle_x, right_ankle_y = keypoint[get_keypoint.RIGHT_ANKLE] return [ nose_x, nose_y, left_shoulder_x, left_shoulder_y, right_shoulder_x, right_shoulder_y, left_elbow_x, left_elbow_y, right_elbow_x, right_elbow_y, left_wrist_x, left_wrist_y, right_wrist_x, right_wrist_y, left_hip_x, left_hip_y, right_hip_x, right_hip_y, left_knee_x, left_knee_y, right_knee_x, right_knee_y, left_ankle_x, left_ankle_y, right_ankle_x, right_ankle_y ] class Helmet: def __init__(self): self.flag = False def getflag(self,det,persondet,annotator,fence=0,point=None,names=None,rname=None,num=1): #print(type(det)) self.flag = False for *xyxy, conf, cls in reversed(det): c = int(cls) labelname = names[c] if labelname in rname: if fence == 1: pointa = strtolstl(point) for poi in pointa: p1 = (int(xyxy[0].cpu().item() + (xyxy[2].cpu().item() - xyxy[0].cpu().item()) / 3), xyxy[3].cpu().item()) p2 = (int(xyxy[0].cpu().item() + (xyxy[2].cpu().item() - xyxy[0].cpu().item()) / 3 * 2), xyxy[3].cpu().item()) pt = [p1, p2] inflag = mat.Path(poi).contains_points(pt) if inflag.any(): if persondet is None: self.flag = True # c = int(cls) # integer class # label = f'{self.names[c]} {conf:.2f}' label = None annotator.box_label(xyxy, None, color=(0, 0, 255)) else: for person in persondet: personinflag = mat.Path(person).contains_points(pt) if personinflag.any(): annotator.box_label(xyxy, None, color=(0, 0, 255)) self.flag = True else: if persondet is None: self.flag = True # c = int(cls) # integer class # label = f'{self.names[c]} {conf:.2f}' label = None annotator.box_label(xyxy, None, color=(0, 0, 255)) else: for person in persondet: p1 = (int(xyxy[0].cpu().item() + (xyxy[2].cpu().item() - xyxy[0].cpu().item()) / 3), xyxy[3].cpu().item()) p2 = (int(xyxy[0].cpu().item() + (xyxy[2].cpu().item() - xyxy[0].cpu().item()) / 3 * 2), xyxy[3].cpu().item()) pt = [p1, p2] personinflag = mat.Path(person).contains_points(pt) if personinflag.any(): annotator.box_label(xyxy, None, color=(0, 0, 255)) self.flag = True return self.flag class Uniform: def __init__(self): self.flag = False def getflag(self, det, persondet,annotator, fence=0, point=None, names=None, rname=None,num=1): self.flag = False for *xyxy, conf, cls in reversed(det): c = int(cls) labelname = names[c] if labelname in rname: if fence == 1: pointa = strtolstl(point) for poi in pointa: p1 = (int(xyxy[0].cpu().item() + (xyxy[2].cpu().item() - xyxy[0].cpu().item()) / 3), xyxy[3].cpu().item()) p2 = (int(xyxy[0].cpu().item() + (xyxy[2].cpu().item() - xyxy[0].cpu().item()) / 3 * 2), xyxy[3].cpu().item()) pt = [p1, p2] inflag = mat.Path(poi).contains_points(pt) if inflag.any(): if persondet is None: self.flag = True # c = int(cls) # integer class # label = f'{self.names[c]} {conf:.2f}' label = None annotator.box_label(xyxy, None, color=(0, 0, 255)) else: for person in persondet: p1 = (int(xyxy[0].cpu().item() + (xyxy[2].cpu().item() - xyxy[0].cpu().item()) / 3), xyxy[3].cpu().item()) p2 = ( int(xyxy[0].cpu().item() + (xyxy[2].cpu().item() - xyxy[0].cpu().item()) / 3 * 2), xyxy[3].cpu().item()) pt = [p1, p2] personinflag = mat.Path(person).contains_points(pt) if personinflag.any(): annotator.box_label(xyxy, None, color=(0, 0, 255)) self.flag = True else: if persondet is None: self.flag = True # c = int(cls) # integer class # label = f'{self.names[c]} {conf:.2f}' label = None annotator.box_label(xyxy, None, color=(0, 0, 255)) else: for person in persondet: p1 = (int(xyxy[0].cpu().item() + (xyxy[2].cpu().item() - xyxy[0].cpu().item()) / 3), xyxy[3].cpu().item()) p2 = (int(xyxy[0].cpu().item() + (xyxy[2].cpu().item() - xyxy[0].cpu().item()) / 3 * 2), xyxy[3].cpu().item()) pt = [p1, p2] personinflag = mat.Path(person).contains_points(pt) if personinflag.any(): annotator.box_label(xyxy, None, color=(0, 0, 255)) self.flag = True return self.flag class Fall: def __init__(self): self.flag = False def getflag(self, det, persondet,annotator, fence=0, point=None, names=None, rname=None,num=1): self.flag = False for *xyxy, conf, cls in reversed(det): c = int(cls) labelname = names[c] if labelname in rname: if fence == 1: pointa = strtolstl(point) for poi in pointa: p1 = (int(xyxy[0].cpu().item() + (xyxy[2].cpu().item() - xyxy[0].cpu().item()) / 3), xyxy[3].cpu().item()) p2 = (int(xyxy[0].cpu().item() + (xyxy[2].cpu().item() - xyxy[0].cpu().item()) / 3 * 2), xyxy[3].cpu().item()) pt = [p1, p2] inflag = mat.Path(poi).contains_points(pt) if inflag.any(): if persondet is None: self.flag = True # c = int(cls) # integer class # label = f'{self.names[c]} {conf:.2f}' label = None annotator.box_label(xyxy, None, color=(0, 0, 255)) else: for person in persondet: personinflag = mat.Path(person).contains_points(pt) if personinflag.any(): annotator.box_label(xyxy, None, color=(0, 0, 255)) self.flag = True else: if persondet is None: self.flag = True # c = int(cls) # integer class # label = f'{self.names[c]} {conf:.2f}' label = None annotator.box_label(xyxy, None, color=(0, 0, 255)) else: for person in persondet: p1 = (int(xyxy[0].cpu().item() + (xyxy[2].cpu().item() - xyxy[0].cpu().item()) / 3), xyxy[3].cpu().item()) p2 = (int(xyxy[0].cpu().item() + (xyxy[2].cpu().item() - xyxy[0].cpu().item()) / 3 * 2), xyxy[3].cpu().item()) pt = [p1, p2] personinflag = mat.Path(person).contains_points(pt) if personinflag.any(): annotator.box_label(xyxy, None, color=(0, 0, 255)) self.flag = True return self.flag class newFall: def __init__(self): self.flag = False self.pose_classfier_model = Model(None, 13, 2, 2) self.pose_classfier_model.load_state_dict(torch.load("posefallcls.pt")) self.pose_classfier_model.eval() self.pose_classfier_model.to(0) def getflag(self, det, persondet,annotator, fence=0, point=None, names=None, rname=None,num=1,imshape=None): self.flag = False kptshape = [17,3] pred_kpts = det[:, 6:].view(len(det), *kptshape) if len(det) else det[:, 6:] pred_kpts = ops.scale_coords(imshape, pred_kpts, annotator.result().shape) results =Results(annotator.result(), path=None, names=names, boxes=det[:, :6], keypoints=pred_kpts) print(results.boxes) if len(point) > 1: self.polygon_areas = [Polygon(strtolstl(point)[0])] print(self.polygon_areas) else: self.polygon_areas = None tmppersondet = [] if self.polygon_areas is not None: for person in persondet: polygon_box = Polygon( [(person[0], person[1]), (person[2], person[3]), (person[4], person[5]), (person[6], person[7])]) intersection_areas = [polygon_area.intersection(polygon_box).area for polygon_area in self.polygon_areas] if sum(intersection_areas) == 0: continue intersection_areas_ratio = sorted( [intersection_area / polygon_box.area for intersection_area in intersection_areas]) # print(intersection_areas_ratio) if intersection_areas_ratio[-1] < 0.9: continue tmppersondet.append(person) persondet = tmppersondet boxes = results.boxes.xyxy.cpu().numpy().tolist() confs = results.boxes.conf.cpu().numpy().tolist() all_keypoints = results.keypoints.data.cpu().numpy().tolist() pose_classfier_results = [] for box, conf, keypoints in zip(boxes, confs, all_keypoints): iouflag = False for personbox in persondet: iou, _ = compute_IOU(box, personbox) print(f'judgeiou = {iou}') if iou > 0.5: iouflag = True break; if not iouflag: break x1, y1, x2, y2 = box x, y, w, h = x1, y1, x2 - x1, y2 - y1 n_keypoints = [[(kp[0] - x) / w - 0.5, (kp[1] - y) / h - 0.5] if kp[0] > 0 and kp[1] > 0 else kp[:2] for kp in keypoints] n_keypoints = extract_keypoint(self.get_keypoint, n_keypoints) if n_keypoints[-12:].count(0) >= 2 * 2: continue if n_keypoints.count(0) >= 4 * 2: continue if w < h: continue pose_data = torch.Tensor([n_keypoints]).to(self.device) pose_data = pose_data.reshape(1, 13, 2) with torch.no_grad(): p = self.pose_classfier_model(pose_data) prob = F.softmax(p) index = prob.argmax() if index == 0: score = float(prob[0][index].cpu().numpy()) pose_classfier_results.append(box) if len(pose_classfier_results)>0: for xyxy in pose_classfier_results: annotator.box_label(xyxy, None, color=(0, 0, 255)) self.flag = True return self.flag class Personcount: def __init__(self): self.flag = False def getflag(self, det,persondet, annotator, fence=0, point=None, names=None, rname=None,num=1): self.flag = False detnum = 0 for *xyxy, conf, cls in reversed(det): c = int(cls) labelname = names[c] if labelname in rname: if fence == 1: pointa = strtolstl(point) for poi in pointa: p1 = (int(xyxy[0].cpu().item() + (xyxy[2].cpu().item() - xyxy[0].cpu().item()) / 3), xyxy[3].cpu().item()) p2 = (int(xyxy[0].cpu().item() + (xyxy[2].cpu().item() - xyxy[0].cpu().item()) / 3 * 2), xyxy[3].cpu().item()) pt = [p1, p2] inflag = mat.Path(poi).contains_points(pt) if inflag.any(): if persondet is None: #self.flag = True detnum = detnum+1 # c = int(cls) # integer class # label = f'{self.names[c]} {conf:.2f}' label = None annotator.box_label(xyxy, None, color=(0, 0, 255)) else: for person in persondet: personinflag = mat.Path(person).contains_points(pt) if personinflag.any(): detnum = detnum+1 annotator.box_label(xyxy, None, color=(0, 0, 255)) #self.flag = True else: if persondet is None: #self.flag = True detnum = detnum+1 # c = int(cls) # integer class # label = f'{self.names[c]} {conf:.2f}' label = None annotator.box_label(xyxy, None, color=(0, 0, 255)) else: for person in persondet: p1 = (int(xyxy[0].cpu().item() + (xyxy[2].cpu().item() - xyxy[0].cpu().item()) / 3), xyxy[3].cpu().item()) p2 = (int(xyxy[0].cpu().item() + (xyxy[2].cpu().item() - xyxy[0].cpu().item()) / 3 * 2), xyxy[3].cpu().item()) pt = [p1, p2] personinflag = mat.Path(person).contains_points(pt) if personinflag.any(): detnum = detnum+1 annotator.box_label(xyxy, None, color=(0, 0, 255)) #self.flag = True if detnum >= num: self.flag = True return self.flag class Arm: def __init__(self): self.flag = False def getflag(self, det, persondet,annotator, fence=0, point=None, names=None, rname=None,num=1): self.flag = False for *xyxy, conf, cls in reversed(det): c = int(cls) labelname = names[c] if labelname in rname: if fence == 1: pointa = strtolstl(point) for poi in pointa: p1 = (int(xyxy[0].cpu().item() + (xyxy[2].cpu().item() - xyxy[0].cpu().item()) / 3), xyxy[3].cpu().item()) p2 = (int(xyxy[0].cpu().item() + (xyxy[2].cpu().item() - xyxy[0].cpu().item()) / 3 * 2), xyxy[3].cpu().item()) pt = [p1, p2] inflag = mat.Path(poi).contains_points(pt) if inflag.any(): if persondet is None: self.flag = True # c = int(cls) # integer class # label = f'{self.names[c]} {conf:.2f}' label = None annotator.box_label(xyxy, None, color=(0,0,255)) else: for person in persondet: personinflag = mat.Path(person).contains_points(pt) if personinflag.any(): annotator.box_label(xyxy, None, color=(0, 0, 255)) self.flag = True else: if persondet is None: self.flag = True # c = int(cls) # integer class # label = f'{self.names[c]} {conf:.2f}' label = None annotator.box_label(xyxy, None, color=(0, 0, 255)) else: for person in persondet: p1 = (int(xyxy[0].cpu().item() + (xyxy[2].cpu().item() - xyxy[0].cpu().item()) / 3), xyxy[3].cpu().item()) p2 = (int(xyxy[0].cpu().item() + (xyxy[2].cpu().item() - xyxy[0].cpu().item()) / 3 * 2), xyxy[3].cpu().item()) pt = [p1, p2] personinflag = mat.Path(np.array(person).reshape(-1,2)).contains_points(pt) if personinflag.any(): annotator.box_label(xyxy, None, color=(0, 0, 255)) self.flag = True return self.flag class Bag: def __init__(self): self.flag = False def getflag(self, det, persondet,annotator, fence=0, point=None, names=None, rname=None,num=1): self.flag = False for *xyxy, conf, cls in reversed(det): c = int(cls) labelname = names[c] if labelname in rname: if fence == 1: pointa = strtolstl(point) for poi in pointa: p1 = (int(xyxy[0].cpu().item() + (xyxy[2].cpu().item() - xyxy[0].cpu().item()) / 3), xyxy[3].cpu().item()) p2 = (int(xyxy[0].cpu().item() + (xyxy[2].cpu().item() - xyxy[0].cpu().item()) / 3 * 2), xyxy[3].cpu().item()) pt = [p1, p2] inflag = mat.Path(poi).contains_points(pt) if inflag.any(): if persondet is None: self.flag = True # c = int(cls) # integer class # label = f'{self.names[c]} {conf:.2f}' label = None annotator.box_label(xyxy, None, color=(0, 0, 255)) else: for person in persondet: personinflag = mat.Path(person).contains_points(pt) if personinflag.any(): annotator.box_label(xyxy, None, color=(0, 0, 255)) self.flag = True else: if persondet is None: self.flag = True # c = int(cls) # integer class # label = f'{self.names[c]} {conf:.2f}' label = None annotator.box_label(xyxy, None, color=(0, 0, 255)) else: for person in persondet: p1 = (int(xyxy[0].cpu().item() + (xyxy[2].cpu().item() - xyxy[0].cpu().item()) / 3), xyxy[3].cpu().item()) p2 = (int(xyxy[0].cpu().item() + (xyxy[2].cpu().item() - xyxy[0].cpu().item()) / 3 * 2), xyxy[3].cpu().item()) pt = [p1, p2] personinflag = mat.Path(person).contains_points(pt) if personinflag.any(): annotator.box_label(xyxy, None, color=(0, 0, 255)) self.flag = True return self.flag class Cross: def __init__(self): self.flag = False def getflag(self, det, persondet,annotator, fence=0, point=None, names=None, rname=None,num=1): self.flag = False detnum = 0 for *xyxy, conf, cls in reversed(det): c = int(cls) labelname = names[c] if labelname in rname: if fence == 1: pointa = strtolstl(point) for poi in pointa: p1 = (int(xyxy[0].cpu().item() + (xyxy[2].cpu().item() - xyxy[0].cpu().item()) / 3), xyxy[3].cpu().item()) p2 = (int(xyxy[0].cpu().item() + (xyxy[2].cpu().item() - xyxy[0].cpu().item()) / 3 * 2), xyxy[3].cpu().item()) pt = [p1, p2] inflag = mat.Path(poi).contains_points(pt) if inflag.any(): if persondet is None: detnum = detnum+1 self.flag = True # c = int(cls) # integer class # label = f'{self.names[c]} {conf:.2f}' label = None annotator.box_label(xyxy, None, color=(0, 0, 255)) else: for person in persondet: personinflag = mat.Path(person).contains_points(pt) if personinflag.any(): detnum = detnum+1 annotator.box_label(xyxy, None, color=(0, 0, 255)) self.flag = True else: if persondet is None: self.flag = True # c = int(cls) # integer class # label = f'{self.names[c]} {conf:.2f}' detnum = detnum+1 label = None annotator.box_label(xyxy, None, color=(0, 0, 255)) else: for person in persondet: p1 = (int(xyxy[0].cpu().item() + (xyxy[2].cpu().item() - xyxy[0].cpu().item()) / 3), xyxy[3].cpu().item()) p2 = (int(xyxy[0].cpu().item() + (xyxy[2].cpu().item() - xyxy[0].cpu().item()) / 3 * 2), xyxy[3].cpu().item()) pt = [p1, p2] personinflag = mat.Path(person).contains_points(pt) if personinflag.any(): detnum = detnum +1 annotator.box_label(xyxy, None, color=(0, 0, 255)) self.flag = True return self.flag class Extinguisher: def __init__(self): self.flag = False def getflag(self, det, persondet,annotator, fence=0, point=None, names=None, rname=None,num=1): self.flag = False for *xyxy, conf, cls in reversed(det): c = int(cls) labelname = names[c] if labelname in rname: if fence == 1: pointa = strtolstl(point) for poi in pointa: p1 = (int(xyxy[0].cpu().item() + (xyxy[2].cpu().item() - xyxy[0].cpu().item()) / 3), xyxy[3].cpu().item()) p2 = (int(xyxy[0].cpu().item() + (xyxy[2].cpu().item() - xyxy[0].cpu().item()) / 3 * 2), xyxy[3].cpu().item()) pt = [p1, p2] inflag = mat.Path(poi).contains_points(pt) if inflag.any(): if persondet is None: self.flag = True # c = int(cls) # integer class # label = f'{self.names[c]} {conf:.2f}' label = None annotator.box_label(xyxy, None, color=(0, 0, 255)) else: for person in persondet: personinflag = mat.Path(person).contains_points(pt) if personinflag.any(): annotator.box_label(xyxy, None, color=(0, 0, 255)) self.flag = True else: if persondet is None: self.flag = True # c = int(cls) # integer class # label = f'{self.names[c]} {conf:.2f}' label = None annotator.box_label(xyxy, None, color=(0, 0, 255)) else: for person in persondet: p1 = (int(xyxy[0].cpu().item() + (xyxy[2].cpu().item() - xyxy[0].cpu().item()) / 3), xyxy[3].cpu().item()) p2 = (int(xyxy[0].cpu().item() + (xyxy[2].cpu().item() - xyxy[0].cpu().item()) / 3 * 2), xyxy[3].cpu().item()) pt = [p1, p2] personinflag = mat.Path(person).contains_points(pt) if personinflag.any(): annotator.box_label(xyxy, None, color=(0, 0, 255)) self.flag = True return self.flag def calculate_iou(box1, box2): # 计算交集区域 也就是找到次左上 次右下 x1_int = max(box1[0], box2[0]) y1_int = max(box1[1], box2[1]) x2_int = min(box1[2], box2[2]) y2_int = min(box1[3], box2[3]) # 交集区域的宽高 width_int = max(0, x2_int - x1_int) height_int = max(0, y2_int - y1_int) area_int = width_int * height_int # 计算两个框的面积 area_box1 = (box1[2] - box1[0]) * (box1[3] - box1[1]) area_box2 = (box2[2] - box2[0]) * (box2[3] - box1[1]) # 计算并集面积 area_union = area_box1 + area_box2 - area_int # 计算IoU iou = area_int / area_union if area_union > 0 else 0 return iou def check_cls4_overlap(person_box, target_box, cls4_boxes): """ 检查是否有 cls4 物体在目标框与人物框交集区域内，并计算交集面积与 cls4 物体面积的重叠比率。 返回重叠比率最大值以及其对应的交集区域坐标和 cls4 框。 参数: person_box (tuple): 人物框 (x_min, y_min, x_max, y_max) target_box (tuple): 目标框 (x_min, y_min, x_max, y_max) cls4_boxes (list): cls4 物体框的列表，每个框为 (x_min, y_min, x_max, y_max) 返回: tuple: - 最大的重叠比率 (float)，如果没有符合条件的物体则为 None。 - 最大重叠比率对应的交集区域的坐标 (inter_x1, inter_y1)。 - 对应的 cls4 框 (tuple)，如果没有符合条件的物体则为 None。 """ # 计算目标框与人物框的交集区域 min_x = max(person_box[0], target_box[0]) min_y = max(person_box[1], target_box[1]) max_x = min(person_box[2], target_box[2]) max_y = min(person_box[3], target_box[3]) target_area = (min_x, min_y, max_x, max_y) # 交集区域 max_overlap_ratio = -1 # 用于存储最大的重叠比率 这设置为-1有助于区别是否计算了 best_inter_x1 = 0 # 用于存储最大重叠比率对应的交集坐标 best_inter_y1 = 0 best_cls4_box = [] # 用于存储对应的 cls4 框 不能初始化为整数0 整数不可迭代 后序不能添加 # 检查是否有 cls4 物体在交集区域内 for cls4_box in cls4_boxes: # 判断 cls4 框是否与交集区域有交集 inter_x1 = max(cls4_box[0], target_area[0]) inter_y1 = max(cls4_box[1], target_area[1]) inter_x2 = min(cls4_box[2], target_area[2]) inter_y2 = min(cls4_box[3], target_area[3]) if inter_x1 < inter_x2 and inter_y1 < inter_y2: # 计算交集区域的面积 intersection_area = (inter_x2 - inter_x1) * (inter_y2 - inter_y1) #print("交集的面积", intersection_area) # 计算 cls4_box 的面积 cls4_area = (cls4_box[2] - cls4_box[0]) * (cls4_box[3] - cls4_box[1]) #print("cls4_area的面积是", cls4_area) # 计算交集区域占 cls4_box 面积的比例 overlap_ratio = intersection_area / cls4_area # 更新最大重叠比率及其相关信息 if max_overlap_ratio == 0 or overlap_ratio > max_overlap_ratio: max_overlap_ratio = overlap_ratio best_inter_x1 = inter_x1 best_inter_y1 = inter_y1 best_cls4_box = cls4_box return max_overlap_ratio, (best_inter_x1, best_inter_y1), best_cls4_box class Persontre1: def __init__(self): self.flag = False self.classifier_model = torch.load('/home/h3c/yolo/persontrecls.pt') def getflag(self, det, persondet,annotator, fence=0, point=None, names=None, rname=None,num=1,im0=None): self.flag = False dirp = {} dirf = {} for *xyxy, conf, cls in reversed(det): c = int(cls) labelname = names[c] if labelname in rname: if fence == 1: pointa = strtolstl(point) for poi in pointa: p1 = (int(xyxy[0].cpu().item() + (xyxy[2].cpu().item() - xyxy[0].cpu().item()) / 3), xyxy[3].cpu().item()) p2 = (int(xyxy[0].cpu().item() + (xyxy[2].cpu().item() - xyxy[0].cpu().item()) / 3 * 2), xyxy[3].cpu().item()) pt = [p1, p2] inflag = mat.Path(poi).contains_points(pt) if inflag.any(): if persondet is None: self.flag = True # c = int(cls) # integer class # label = f'{self.names[c]} {conf:.2f}' label = None if c==0: dirp.setdefault(0,[]) dirp[0].append(xyxy) elif c in [1,2]: dirp.setdefault(1,[]) dirp[1].append(xyxy) dirf.setdefault(1,[]) dirf[1].append(xyxy) elif c==3: dirp.setdefault(1,[]) dirp[1].append(xyxy) elif c==4: dirf.setdefault(0,[]) dirf[0].append(xyxy) #annotator.box_label(xyxy, None, color=(0, 0, 255)) else: for person in persondet: personinflag = mat.Path(person).contains_points(pt) if personinflag.any(): #annotator.box_label(xyxy, None, color=(0, 0, 255)) #self.flag = True if c==0: dirp.setdefault(0,[]) dirp[0].append(xyxy) elif c in [1,2]: dirp.setdefault(1,[]) dirp[1].append(xyxy) dirf.setdefault(1,[]) dirf[1].append(xyxy) elif c==3: dirp.setdefault(1,[]) dirp[1].append(xyxy) elif c==4: dirf.setdefault(0,[]) dirf[0].append(xyxy) else: if persondet is None: #self.flag = True # c = int(cls) # integer class # label = f'{self.names[c]} {conf:.2f}' #label = None #annotator.box_label(xyxy, None, color=(0, 0, 255)) if c==0: dirp.setdefault(0,[]) dirp[0].append(xyxy) elif c in [1,2]: dirp.setdefault(1,[]) dirp[1].append(xyxy) dirf.setdefault(1,[]) dirf[1].append(xyxy) elif c==3: dirp.setdefault(1,[]) dirp[1].append(xyxy) elif c==4: dirf.setdefault(0,[]) dirf[0].append(xyxy) else: for person in persondet: p1 = (int(xyxy[0].cpu().item() + (xyxy[2].cpu().item() - xyxy[0].cpu().item()) / 3), xyxy[3].cpu().item()) p2 = (int(xyxy[0].cpu().item() + (xyxy[2].cpu().item() - xyxy[0].cpu().item()) / 3 * 2), xyxy[3].cpu().item()) pt = [p1, p2] personinflag = mat.Path(person).contains_points(pt) if personinflag.any(): #annotator.box_label(xyxy, None, color=(0, 0, 255)) #self.flag = True if c==0: dirp.setdefault(0,[]) dirp[0].append(xyxy) elif c in [1,2]: dirp.setdefault(1,[]) dirp[1].append(xyxy) dirf.setdefault(1,[]) dirf[1].append(xyxy) elif c==3: dirp.setdefault(1,[]) dirp[1].append(xyxy) elif c==4: dirf.setdefault(0,[]) dirf[0].append(xyxy) if len(dirp.keys()) == 2: claimg = [] clapoint = [] for person in dirp[0]: for other in dirp[1]: iou, newxyxy = compute_IOU(person, other) if iou>0.1: print(newxyxy) imgtmp = im0[int(newxyxy[1]):int(newxyxy[3]),int(newxyxy[0]):int(newxyxy[2])] imgtmp = imgtmp[...,::-1] imgtmp = Image.fromarray(imgtmp) imgten1 = test(imgtmp) claimg.append(imgten1) clapoint.append((newxyxy)) result = clapre(self.classifier_model,claimg,clapoint) #imgten = imgten1[None] #imgten = imgten.to(0) #result = modelcla(imgten) #result = F.softmax(result, dim=1) #cla = result.argmax(1) if result is not None: self.flag = True for res in result: print(res) annotator.box_label(res, None, color=(0,0,255)) if len(dirp.keys()) == 2: claimg = [] clapoint = [] for person in dirp[0]: for other in dirp[1]: iou, newxyxy = compute_IOU(person, other) if iou>0.1: print(newxyxy) imgtmp = im0[int(newxyxy[1]):int(newxyxy[3]),int(newxyxy[0]):int(newxyxy[2])] imgtmp = imgtmp[...,::-1] imgtmp = Image.fromarray(imgtmp) imgten1 = test(imgtmp) claimg.append(imgten1) clapoint.append((newxyxy)) result = clapre(self.classifier_model,claimg,clapoint) #imgten = imgten1[None] #imgten = imgten.to(0) #result = modelcla(imgten) #result = F.softmax(result, dim=1) #cla = result.argmax(1) if result is not None: self.flag = True for res in result: print(res) annotator.box_label(res, None, color=(0,0,255)) return self.flag class Persontree: def __init__(self): self.flag = False self.classifier_model = torch.load('/home/h3c/yolo/persontrecls.pt') def getflag(self, det, persondet,annotator, fence=0, point=None, names=None, rname=None,num=1,im0=None): self.flag = False target_classes = [1, 2, 3] results = Results(annotator.result(),path=None,names=names,boxes=det) boxes = results.boxes person_boxes = [] target_boxes = [] cls4_boxes = [] # 处理检测结果 for i in range(len(boxes)): cls = int(boxes.cls[i].item()) con = boxes.conf[i].item() if cls == 0 and con > 0.1: # 如果是 "person" 类别 x1, y1, x2, y2 = boxes.xyxy[i].tolist() person_boxes.append([x1, y1, x2, y2]) if cls in target_classes and con > 0.1: # 目标类别（bag, box, cart） x1, y1, x2, y2 = boxes.xyxy[i].tolist() target_boxes.append([x1, y1, x2, y2]) if cls == 4 and con > 0.1: # 如果是 "cls 4" 类别 x1, y1, x2, y2 = boxes.xyxy[i].tolist() cls4_boxes.append([x1, y1, x2, y2]) # 如果检测到 "person" 类别和目标框，计算IoU if person_boxes and target_boxes: for i, person_box in enumerate(person_boxes): person_center_y = (person_box[1] + person_box[3]) / 2 for j, target_box in enumerate(target_boxes): target_center_y = (target_box[1] + target_box[3]) / 2 # 判断目标框的中心点是否在person框的下方 if target_center_y + 20 > person_center_y: # 根据需要调整此阈值 iou = calculate_iou(person_box, target_box) if iou > 0: # IoU大于0，进入新的判断 # 创建一个包围person和target框的区域 本来思路是判断脚是否在这个大框框里面 但是这个不合理 应该判断脚是不是在这个交集里面，再检测脚和人与物交集有没有交集 min_x = max(person_box[0], target_box[0]) min_y = max(person_box[1], target_box[1]) max_x = min(person_box[2], target_box[2]) max_y = min(person_box[3], target_box[3]) target_area = (min_x, min_y, max_x, max_y) # 检查是否有cls 4物体在这个区域内 for cls4_box in cls4_boxes: #比较巧妙 提前存储这个变量 然后检测这个物体是不是在这里面 # 判断cls 4框是否与这个区域有交集 inter_x1 = max(cls4_box[0], target_area[0]) inter_y1 = max(cls4_box[1], target_area[1]) inter_x2 = min(cls4_box[2], target_area[2]) inter_y2 = min(cls4_box[3], target_area[3]) if inter_x1 < inter_x2 and inter_y1 < inter_y2: # 计算交集区域的面积 intersection_area = (inter_x2 - inter_x1) * (inter_y2 - inter_y1) # 计算 cls4_box 的面积 cls4_area = (cls4_box[2] - cls4_box[0]) * (cls4_box[3] - cls4_box[1]) # 计算交集区域占 cls4_box 面积的比例 overlap_ratio = intersection_area / cls4_area if overlap_ratio > 0.5: self.flag = True return self.flag class Persontre: def __init__(self): self.flag = False self.classifier_model = torch.load('/home/h3c/yolo/persontrecls.pt') def getflag(self, det, persondet,annotator, fence=0, point=None, names=None, rname=None,num=1,im0=None): self.flag = False cls3 =3 #对cart单独一类 cls4 = 4 # 新增加的类别编号（假设为 4） target_classes = [1, 2, 3] results = Results(annotator.result(),path=None,names=names,boxes=det) boxes = results.boxes person_boxes = [] # 人 车子 脚 以及 指定区域内的脚 cls3_boxes = [] cls4_boxes = [] best_cls4_box = [] # 处理检测结果 for i in range(len(boxes)): cls = int(boxes.cls[i].item()) con = boxes.conf[i].item() if cls == 0 and con > 0.1: # 如果是 "person" 类别 先把对应类别添加上去 先把每一个识别物体输入上去 x1, y1, x2, y2 = boxes.xyxy[i].tolist() person_boxes.append([x1, y1, x2, y2]) if cls == cls3 and con > 0.1: # 目标类别 cart x1, y1, x2, y2 = boxes.xyxy[i].tolist() cls3_boxes.append([x1, y1, x2, y2]) if cls == cls4 and con > 0.1: # 如果是 "cls 4" 类别 x1, y1, x2, y2 = boxes.xyxy[i].tolist() cls4_boxes.append([x1, y1, x2, y2]) # 如果检测到 "person" 类别和目标框，计算IoU if person_boxes and cls3_boxes: # 先查看这俩个类别有没有 for i, person_box in enumerate(person_boxes): for j, cls3_box in enumerate(cls3_boxes): # 判断目标框的中心点是否在person框的下方 iou = calculate_iou(person_box, cls3_box) # 踩的逻辑 一个是要求车子和人有交并比 其次判断脚是不是在这个交并比内 并且比重不能太低 if iou > 0.2: # 用来判断坐 如果iou足够高 就视为是坐 # 加载图像并绘制标注 # 看看这个值大概有多少 好衡量一下 #print(f"这个iou大于0.2的 这是第{frame_idx},iou是{iou}") # 写入介绍 annotator.box_label(person_box, None, color=(0, 0, 255)) annotator.box_label(cls3_box, None, color=(0, 0, 255)) self.flag = True #print("这是坐",f"{saved_image_path}") elif iou > 0 : # IoU大于0，进入新的判断 代表不是坐 # 创建一个包围person和target框的区域 本来思路是判断脚是否在这个大框框里面 但是这个不合理 应该判断脚是不是在这个交集里面，再检测脚和人与物交集有没有交集 overlap_ratio, inter_xy, best_cls4_box1 = check_cls4_overlap(person_box, cls3_box,cls4_boxes) best_cls4_box.extend(best_cls4_box1) #拿到这四个点的坐标 inter_x1, inter_y1 = inter_xy # 如果占比大于某个阈值，可以执行进一步的操作 if overlap_ratio > 0.2: annotator.box_label(person_box, None, color=(0, 0, 255)) annotator.box_label(cls3_box, None, color=(0, 0, 255)) annotator.box_label(best_cls4_box, None, color=(0, 0, 255)) self.flag = True class newPersontre: def __init__(self): self.flag = False self.classifier_model = torch.load('/home/h3c/yolo/persontrecls.pt') def getflag(self, det, persondet, annotator, fence=0, point=None, names=None, rname=None, num=1, im0=None): self.flag = False cls3 = 1 # 对cart单独一类 # cls4 = 4 # 新增加的类别编号（假设为 4） # target_classes = [1, 2, 3] results = Results(annotator.result(), path=None, names=names, boxes=det) boxes = results.boxes print(boxes) person_boxes = [] # 人 车子 脚 以及 指定区域内的脚 cls3_boxes = [] # cls4_boxes = [] # best_cls4_box = [] # 处理检测结果 for i in range(len(boxes)): cls = int(boxes.cls[i].item()) con = boxes.conf[i].item() if cls == 0 and con > 0.1: # 如果是 "person" 类别 先把对应类别添加上去 先把每一个识别物体输入上去 x1, y1, x2, y2 = boxes.xyxy[i].tolist() person_boxes.append([x1, y1, x2, y2]) if cls == cls3 and con > 0.1: # 目标类别 cart x1, y1, x2, y2 = boxes.xyxy[i].tolist() cls3_boxes.append([x1, y1, x2, y2]) print(person_boxes) print(cls3_boxes) # 如果检测到 "person" 类别和目标框，计算IoU if person_boxes and cls3_boxes: # 先查看这俩个类别有没有 for i, person_box in enumerate(person_boxes): for j, cls3_box in enumerate(cls3_boxes): # 判断是否存在交集 而且不能低 iou = calculate_iou(person_box, cls3_box) print(f"IOU的大小这是第", iou) if iou > 0.15: # 用来判断坐 如果iou足够高 就视为是坐 # 加载图像并绘制标注 # 看看这个值大概有多少 好衡量一下 print(f"这个iou大于0.2的 这是第,iou是{iou}") # 写入介绍 # img = Image.fromarray(cv2.cvtColor(frame, # cv2.COLOR_BGR2RGB)) # 注意 cap.read 获取到的帧是numpy形式 不能直接配合Draw函数 需要转换 # 创建绘图上下文 # draw = ImageDraw.Draw(img) # 继续进行绘制操作 # 绘制 "person" 类别框 annotator.box_label(person_box, None, color=(0, 0, 255)) # 绘制目标框（bag, box, cart） annotator.box_label(cls3_box, None, color=(0, 0, 255)) self.flag = True return self.flag class Danager: def __init__(self): self.flag = False def getflag(self, det, persondet,annotator, fence=0, point=None, names=None, rname=None,num=1): self.flag = False for *xyxy, conf, cls in reversed(det): c = int(cls) labelname = names[c] if labelname in rname: if fence == 1: pointa = strtolstl(point) for poi in pointa: p1 = (int(xyxy[0].cpu().item() + (xyxy[2].cpu().item() - xyxy[0].cpu().item()) / 2), int(xyxy[1].cpu().item() + (xyxy[3].cpu().item() - xyxy[1].cpu().item()) / 2)) #p2 = (int(xyxy[0].cpu().item() + (xyxy[2].cpu().item() - xyxy[0].cpu().item()) / 3 * 2), #xyxy[3].cpu().item()) pt = [p1] inflag = mat.Path(poi).contains_points(pt) if inflag.any(): if persondet is None: self.flag = True # c = int(cls) # integer class # label = f'{self.names[c]} {conf:.2f}' label = None annotator.box_label(xyxy, None, color=(0, 0, 255)) else: for person in persondet: personinflag = mat.Path(person).contains_points(pt) if personinflag.any(): annotator.box_label(xyxy, None, color=(0, 0, 255)) self.flag = True else: if persondet is None: self.flag = True # c = int(cls) # integer class # label = f'{self.names[c]} {conf:.2f}' label = None annotator.box_label(xyxy, None, color=(0, 0, 255)) else: for person in persondet: p1 = (int(xyxy[0].cpu().item() + (xyxy[2].cpu().item() - xyxy[0].cpu().item()) / 3), xyxy[3].cpu().item()) p2 = (int(xyxy[0].cpu().item() + (xyxy[2].cpu().item() - xyxy[0].cpu().item()) / 3 * 2), xyxy[3].cpu().item()) pt = [p1, p2] personinflag = mat.Path(person).contains_points(pt) if personinflag.any(): annotator.box_label(xyxy, None, color=(0, 0, 255)) self.flag = True return self.flag class CarHelmetBelt: def __init__(self): self.flag = False def selectNoBeltPerson(self, person_objs, belt_objs): objs = [] polygon_person = [Polygon( [(left, top), (right, top), (right, bottom), (left, bottom)]) for left, top, right, bottom, _, _ in person_objs] polygon_belt = [Polygon( [(left, top), (right, top), (right, bottom), (left, bottom)]) for left, top, right, bottom, _, _ in belt_objs] for person_obj, person in zip(person_objs, polygon_person): with_belt = False for belt in polygon_belt: if person.intersection(belt).area / belt.area > 0.5: with_belt = True break if not with_belt: objs.append(person_obj) return objs def selectWithPersonHead(self, person_objs, head_objs): objs = [] polygon_person = [Polygon( [(left, top), (right, top), (right, top + (bottom - top)/2), (left, top + (bottom - top)/2)]) for left, top, right, bottom, _, _ in person_objs] polygon_head = [Polygon( [(left, top), (right, top), (right, bottom), (left, bottom)]) for left, top, right, bottom, _, _ in head_objs] for head_obj, head in zip(head_objs, polygon_head): with_person = False for person in polygon_person: print('head') if person.intersection(head).area / head.area > 0.5: with_person = True break if with_person: objs.append(head_obj) return objs def getflag(self, det, persondet,annotator, fence=0, point=None, names=None, rname=None,num=1): self.flag = False results = Results(annotator.result(),path=None,names=names,boxes=det) person_objs = [] head_objs = [] belt_objs = [] self.polygon_areas = [Polygon(strtolstl(point)[0])] for result in results: boxes = result.boxes for box in boxes: #print(box.conf.cpu()) left, top, right, bottom = box.xyxy.cpu().numpy().tolist()[0] polygon_box = Polygon([(left, top), (right, top), (right, bottom), (left, bottom)]) intersection_areas = [polygon_area.intersection(polygon_box).area for polygon_area in self.polygon_areas ] if sum(intersection_areas) == 0: continue intersection_areas_ratio = sorted([intersection_area / polygon_box.area for intersection_area in intersection_areas]) if intersection_areas_ratio[-1] < 0.9: continue conf = box.conf.cpu().numpy().tolist()[0] cls = box.cls.cpu().numpy().tolist()[0] if cls == 0: person_objs.append([left, top, right, bottom, conf, names[cls]]) elif cls == 1: head_objs.append([left, top, right, bottom, conf, names[cls]]) elif cls == 3: belt_objs.append([left, top, right, bottom, conf, names[cls]]) #print(head_objs) #print(person_objs) illegal_objs = self.selectNoBeltPerson(person_objs, belt_objs) + self.selectWithPersonHead(person_objs, head_objs) if len(illegal_objs)>0: for obj in illegal_objs: annotator.box_label(obj[:4], None, color=(0, 0, 255)) self.flag = True return self.flag class newUniformi1: def __init__(self): self.flag = False def selectNoUniformPerson(self, person_objs, uniform_objs): objs = [] print(person_objs) print(uniform_objs) polygon_person = [Polygon( [(left, top), (right, top), (right, bottom), (left, bottom)]) for left, top, right, bottom, _, _ in person_objs] polygon_uniform = [Polygon( [(left, top), (right, top), (right, bottom), (left, bottom)]) for left, top, right, bottom, _, _ in uniform_objs] for person_obj, person in zip(person_objs, polygon_person): with_uniform = False for uniform in polygon_uniform: if person.intersection(uniform).area / uniform.area > 0.3: with_uniform = True break if not with_uniform: print(f'illperson_obj {person_obj} illpolygon_uniform {polygon_uniform}') objs.append(person_obj) return objs def getflag(self, det, persondet,annotator, fence=0, point=None, names=None, rname=None,num=1): self.flag = False results = Results(annotator.result(),path=None,names=names,boxes=det) print(results.boxes) person_objs = [] uniform_objs = [] #belt_objs = [] if len(point)>1: self.polygon_areas = [Polygon(strtolstl(point)[0])] else: self.polygon_areas = None if persondet is not None: if self.polygon_areas is not None: for left, top, right,top,right, bottom,left,bottom in persondet: polygon_box = Polygon([(left, top), (right, top), (right, bottom), (left, bottom)]) intersection_areas = [polygon_area.intersection(polygon_box).area for polygon_area in self.polygon_areas] if sum(intersection_areas) == 0: continue intersection_areas_ratio = sorted( [intersection_area / polygon_box.area for intersection_area in intersection_areas]) print(f'inter = {intersection_areas_ratio}') if intersection_areas_ratio[-1] >= 0.3: print("person in check space!!!") width = right - left height = bottom - top print(f'{left}, {top}, {right},{top},{right}, {bottom},{left},{bottom}') print(f'hw = {width* height}') if width * height > 10000: person_objs.append([left, top, right, bottom, 0.25, 'person']) for result in results: boxes = result.boxes for box in boxes: left, top, right, bottom = box.xyxy.cpu().numpy().tolist()[0] # if self.polygon_areas is not None : # if persondet is None and box.cls.cpu().numpy().tolist()[0] == 4: # polygon_box = Polygon([(left, top), (right, top), (right, bottom), (left, bottom)]) # # intersection_areas = [polygon_area.intersection(polygon_box).area for polygon_area in self.polygon_areas ] # if sum(intersection_areas) == 0: # continue # # intersection_areas_ratio = sorted([intersection_area / polygon_box.area for intersection_area in intersection_areas]) # print(f'inter = {intersection_areas_ratio}') # if intersection_areas_ratio[-1] < 0.3: # continue conf = box.conf.cpu().numpy().tolist()[0] cls = box.cls.cpu().numpy().tolist()[0] # if persondet is not None: # if cls == 4 and conf >=0.7: # width = right - left # height = bottom - top # if width * height >10000: # person_objs.append([left, top, right, bottom, conf, names[cls]]) if cls in [1,2]: #width = right - left #height = bottom - top #if width * height >4096: uniform_objs.append([left, top, right, bottom, conf, names[cls]]) print(f'personobjs = {person_objs}') print(f'uniformobjs = {uniform_objs}') illegal_objs = self.selectNoUniformPerson(person_objs, uniform_objs) if len(illegal_objs)>0: for obj in illegal_objs: annotator.box_label(obj[:4], None, color=(0, 0, 255)) self.flag = True return self.flag class newHelmet: def __init__(self): self.flag = False def selectNoHelmetPerson(self, person_objs, head_objs): objs = [] polygon_person = [Polygon( [(left, top), (right, top), (right, bottom), (left, bottom)]) for left, top, right, bottom, _, _ in person_objs] polygon_head = [Polygon( [(left, top), (right, top), (right, bottom), (left, bottom)]) for left, top, right, bottom, _, _ in head_objs] for person_obj, person in zip(person_objs, polygon_person): with_head = False for head in polygon_head: print('head') if person.intersection(head).area / head.area > 0.3: with_head = True break if with_head: objs.append(person_obj) return objs def getflag(self, det, persondet,annotator, fence=0, point=None, names=None, rname=None,num=1): self.flag = False results = Results(annotator.result(),path=None,names=names,boxes=det) person_objs = [] head_objs = [] helmet_objs = [] headtmp = [] #belt_objs = [] if len(point)>1: self.polygon_areas = [Polygon(strtolstl(point)[0])] else: self.polygon_areas = None for result in results: boxes = result.boxes for box in boxes: left, top, right, bottom = box.xyxy.cpu().numpy().tolist()[0] if self.polygon_areas is not None: polygon_box = Polygon([(left, top), (right, top), (right, bottom), (left, bottom)]) intersection_areas = [polygon_area.intersection(polygon_box).area for polygon_area in self.polygon_areas ] if sum(intersection_areas) == 0: continue intersection_areas_ratio = sorted([intersection_area / polygon_box.area for intersection_area in intersection_areas]) if intersection_areas_ratio[-1] < 0.9: continue conf = box.conf.cpu().numpy().tolist()[0] cls = box.cls.cpu().numpy().tolist()[0] #print(f'{conf} {cls}') #print(cls in [5,6,7,8,9,10]) if cls == 4 and conf >=0.7: width = right - left height = bottom - top if width * height >4096: person_objs.append([left, top, right, bottom, conf, names[cls]]) elif cls == 0 and conf >= 0.5: #width = right - left #height = bottom - top #if width * height >4096: print('------') headtmp.append([left, top, right, bottom, conf, names[cls]]) elif cls in [5,6,7,8,9,10]: helmet_objs.append([(left, top), (right, top), (right, bottom), (left, bottom)]) print(f'headtmp= {headtmp}') print(f'helmet_objs = {helmet_objs}') for left, top, right, bottom, conf, name in headtmp: flag = False pt = [(int((left+right)/2),int((top+bottom)/2))] for helmet in helmet_objs: pflag = mat.Path(helmet).contains_points(pt) if pflag.any(): flag = True break if not flag: head_objs.append([left, top, right, bottom, conf, name]) illegal_objs = self.selectNoHelmetPerson(person_objs, head_objs) if len(illegal_objs)>0: for obj in illegal_objs: annotator.box_label(obj[:4], None, color=(0, 0, 255)) self.flag = True return self.flag class newHelmetn: def __init__(self): self.flag = False def selectNoHelmetPerson(self, person_objs, head_objs): objs = [] polygon_person = [Polygon( [(left, top), (right, top), (right, bottom), (left, bottom)]) for left, top, right, bottom, _, _ in person_objs] polygon_head = [Polygon( [(left, top), (right, top), (right, bottom), (left, bottom)]) for left, top, right, bottom, _, _ in head_objs] for person_obj, person in zip(person_objs, polygon_person): with_head = False for head in polygon_head: if person.intersection(head).area / head.area > 0.3: with_head = True break if with_head: objs.append(person_obj) return objs def getflag(self, det, persondet,annotator, fence=0, point=None, names=None, rname=None,num=1): self.flag = False results = Results(annotator.result(),path=None,names=names,boxes=det) person_objs = [] head_objs = [] #belt_objs = [] if len(point)>1: self.polygon_areas = [Polygon(strtolstl(point)[0])] else: self.polygon_areas = None for result in results: boxes = result.boxes for box in boxes: left, top, right, bottom = box.xyxy.cpu().numpy().tolist()[0] if self.polygon_areas is not None: polygon_box = Polygon([(left, top), (right, top), (right, bottom), (left, bottom)]) intersection_areas = [polygon_area.intersection(polygon_box).area for polygon_area in self.polygon_areas ] if sum(intersection_areas) == 0: continue intersection_areas_ratio = sorted([intersection_area / polygon_box.area for intersection_area in intersection_areas]) if intersection_areas_ratio[-1] < 0.9: continue conf = box.conf.cpu().numpy().tolist()[0] cls = box.cls.cpu().numpy().tolist()[0] if cls == 4: width = right - left height = bottom - top if width * height >4096: person_objs.append([left, top, right, bottom, conf, names[cls]]) elif cls == 0: #width = right - left #height = bottom - top #if width * height >4096: head_objs.append([left, top, right, bottom, conf, names[cls]]) illegal_objs = self.selectNoHelmetPerson(person_objs, head_objs) if len(illegal_objs)>0: for obj in illegal_objs: annotator.box_label(obj[:4], None, color=(0, 0, 255)) self.flag = True return self.flag class newUniformt: def __init__(self): self.flag = False def selectNoUniformPerson(self, person_objs, uniform_objs): objs = [] print(person_objs) print(uniform_objs) polygon_person = [Polygon( [(left, top), (right, top), (right, bottom), (left, bottom)]) for left, top, right, bottom, _, _ in person_objs] polygon_uniform = [Polygon( [(left, top), (right, top), (right, bottom), (left, bottom)]) for left, top, right, bottom, _, _ in uniform_objs] for person_obj, person in zip(person_objs, polygon_person): with_uniform = False for uniform in polygon_uniform: if person.intersection(uniform).area / uniform.area > 0.3: with_uniform = True break if not with_uniform: print(f'illperson_obj {person_obj} illpolygon_uniform {polygon_uniform}') objs.append(person_obj) return objs def selectHeadPerson(self, person_objs, head_objs): objs = [] print(person_objs) print(head_objs) #personlist = [] polygon_person = [Polygon( [(left, top), (right, top), (right, bottom), (left, bottom)]) for left, top, right, bottom, _, _ in person_objs] polygon_head = [Polygon( [(left, top), (right, top), (right, bottom), (left, bottom)]) for left, top, right, bottom, _, _ in uniform_objs] for person_obj, person in zip(person_objs, polygon_person): #with_uniform = False for head in polygon_head: if person.intersection(head).area / head.area > 0.3: #with_uniform = True #break #if not with_uniform: # print(f'illperson_obj {person_obj} illpolygon_uniform {polygon_uniform}') objs.append(person_obj) break return objs def getflag(self, det, persondet,annotator, fence=0, point=None, names=None, rname=None,num=1): self.flag = False results = Results(annotator.result(),path=None,names=names,boxes=det) print(results.boxes) person_objs = [] uniform_objs = [] head_objs = [] #belt_objs = [] if len(point)>1: self.polygon_areas = [Polygon(strtolstl(point)[0])] else: self.polygon_areas = None for result in results: boxes = result.boxes for box in boxes: left, top, right, bottom = box.xyxy.cpu().numpy().tolist()[0] if self.polygon_areas is not None: polygon_box = Polygon([(left, top), (right, top), (right, bottom), (left, bottom)]) intersection_areas = [polygon_area.intersection(polygon_box).area for polygon_area in self.polygon_areas ] if sum(intersection_areas) == 0: continue intersection_areas_ratio = sorted([intersection_area / polygon_box.area for intersection_area in intersection_areas]) if intersection_areas_ratio[-1] < 0.9: continue conf = box.conf.cpu().numpy().tolist()[0] cls = box.cls.cpu().numpy().tolist()[0] if cls == 4: width = right - left height = bottom - top if width * height >4096: person_objs.append([left, top, right, bottom, conf, names[cls]]) elif cls in [1,2]: #width = right - left #height = bottom - top #if width * height >4096: uniform_objs.append([left, top, right, bottom, conf, names[cls]]) elif cls != 3: head_objs.append([left, top, right, bottom, conf, names[cls]]) person_objs = self.selectNoUniformPerson(person_objs, head_objs) illegal_objs = self.selectNoUniformPerson(person_objs, uniform_objs) if len(illegal_objs)>0: for obj in illegal_objs: annotator.box_label(obj[:4], None, color=(0, 0, 255)) self.flag = True return self.flag class newUniform: def __init__(self): self.flag = False def selectNoUniformPerson(self, person_objs, uniform_objs): objs = [] print(person_objs) print(uniform_objs) polygon_person = [Polygon( [(left, top), (right, top), (right, bottom), (left, bottom)]) for left, top, right, bottom, _, _ in person_objs] polygon_uniform = [Polygon( [(left, top), (right, top), (right, bottom), (left, bottom)]) for left, top, right, bottom, _, _ in uniform_objs] for person_obj, person in zip(person_objs, polygon_person): with_uniform = False for uniform in polygon_uniform: if person.intersection(uniform).area / uniform.area > 0.3: with_uniform = True break if not with_uniform: print(f'illperson_obj {person_obj} illpolygon_uniform {polygon_uniform}') objs.append(person_obj) return objs def getflag(self, det, persondet,annotator, fence=0, point=None, names=None, rname=None,num=1): self.flag = False results = Results(annotator.result(),path=None,names=names,boxes=det) print(results.boxes) person_objs = [] uniform_objs = [] print(f'persondet = {persondet}') #belt_objs = [] if len(point)>1: self.polygon_areas = [Polygon(strtolstl(point)[0])] print(self.polygon_areas) else: self.polygon_areas = None tmppersondet = [] if self.polygon_areas is not None: for person in persondet: polygon_box = Polygon([(person[0], person[1]), (person[2], person[3]), (person[4], person[5]), (person[6], person[7])]) intersection_areas = [polygon_area.intersection(polygon_box).area for polygon_area in self.polygon_areas ] if sum(intersection_areas) == 0: continue intersection_areas_ratio = sorted([intersection_area / polygon_box.area for intersection_area in intersection_areas]) #print(intersection_areas_ratio) if intersection_areas_ratio[-1] < 0.9: continue tmppersondet.append(person) persondet = tmppersondet print(f'tmppersondet = {tmppersondet}') #persondet = tmppersondet for result in results: boxes = result.boxes for box in boxes: left, top, right, bottom = box.xyxy.cpu().numpy().tolist()[0] conf = box.conf.cpu().numpy().tolist()[0] cls = box.cls.cpu().numpy().tolist()[0] #if self.polygon_areas is not None and cls == 4: # polygon_box = Polygon([(left, top), (right, top), (right, bottom), (left, bottom)]) # intersection_areas = [polygon_area.intersection(polygon_box).area for polygon_area in self.polygon_areas ] # if sum(intersection_areas) == 0: # continue # intersection_areas_ratio = sorted([intersection_area / polygon_box.area for intersection_area in intersection_areas]) # if intersection_areas_ratio[-1] < 0.9: # continue #conf = box.conf.cpu().numpy().tolist()[0] #cls = box.cls.cpu().numpy().tolist()[0] print(conf) print(cls) if cls == 4 and conf >=0.7: width = right - left height = bottom - top print(width*height) if width * height >10000: for person in persondet: #p1 = (int(xyxy[0].cpu().item() + (xyxy[2].cpu().item() - xyxy[0].cpu().item()) / 3), #xyxy[3].cpu().item()) #p2 = (int(xyxy[0].cpu().item() + (xyxy[2].cpu().item() - xyxy[0].cpu().item()) / 3 * 2), # xyxy[3].cpu().item()) personcenterx = int((person[0]+person[2])/2) personcentery = int((person[1]+person[7])/2) p1 = (personcenterx,personcentery) print(p1) pt = [p1] print(f'p1 = {p1}') print(f'person = {person}') personinflag = mat.Path([(left, top), (right, top), (right, bottom), (left, bottom)]).contains_points(pt) if personinflag.any(): person_objs.append([left, top, right, bottom, conf, names[cls]]) elif cls in [1,2]: #width = right - left #height = bottom - top #if width * height >4096: uniform_objs.append([left, top, right, bottom, conf, names[cls]]) illegal_objs = self.selectNoUniformPerson(person_objs, uniform_objs) if len(illegal_objs)>0: for obj in illegal_objs: annotator.box_label(obj[:4], None, color=(0, 0, 255)) self.flag = True return self.flag