import sys, yaml from easydict import EasyDict as edict from concurrent.futures import ThreadPoolExecutor sys.path.extend(['..','../AIlib2' ]) from AI import AI_process,AI_process_forest,get_postProcess_para,get_postProcess_para_dic,ocr_process,AI_det_track,AI_det_track_batch from stdc import stdcModel import cv2,os,time from segutils.segmodel import SegModel from segutils.trafficUtils import tracfficAccidentMixFunction from models.experimental import attempt_load from utils.torch_utils import select_device from utilsK.queRiver import get_labelnames,get_label_arrays,save_problem_images,riverDetSegMixProcess,draw_painting_joint from utilsK.drownUtils import mixDrowing_water_postprocess from ocrUtils.ocrUtils import CTCLabelConverter,AlignCollate from trackUtils.sort import Sort,track_draw_boxAndTrace,track_draw_trace_boxes,moving_average_wang,drawBoxTraceSimplied from obbUtils.load_obb_model import load_model_decoder_OBB from obbUtils.shipUtils import OBB_infer,draw_obb import numpy as np import torch,glob import tensorrt as trt from utilsK.masterUtils import get_needed_objectsIndex from utilsK.noParkingUtils import mixNoParking_road_postprocess from copy import deepcopy from scipy import interpolate #import warnings #warnings.filterwarnings("error") def view_bar(num, total,time1,prefix='prefix'): rate = num / total time_n=time.time() rate_num = int(rate * 30) rate_nums = np.round(rate * 100) r = '\r %s %d / %d [%s%s] %.2f s'%(prefix,num,total, ">" * rate_num, " " * (30 - rate_num), time_n-time1 ) sys.stdout.write(r) sys.stdout.flush() ''' 多线程 ''' def drawAllBox(preds,imgDraw,label_arraylist,rainbows,font): for box in preds: #cls,conf,xyxy = box[0],box[5], box[1:5] #print('#'*20,'line47',box) cls,conf,xyxy = box[5],box[4], box[0:4] ##2023.08.03,修改了格式 #print('#####line46 demo.py:', cls,conf,xyxy, len(label_arraylist),len(rainbows) ) imgDraw = draw_painting_joint(xyxy,imgDraw,label_arraylist[int(cls)],score=conf,color=rainbows[int(cls)%20],font=font,socre_location="leftBottom") return imgDraw def process_v1(frame): #try: time00 = time.time() H,W,C = frame[0][0].shape #frmess---- (im0s,model,segmodel,names,label_arraylist,rainbows,objectPar,digitFont,os.path.basename(imgpath),segPar,mode,postPar) #p_result[1] = draw_painting_joint(xyxy,p_result[1],label_arraylist[int(cls)],score=conf,color=rainbows[int(cls)%20],font=font,socre_location="leftBottom") p_result,timeOut = AI_process(frame[0],frame[1],frame[2],frame[3],frame[4],frame[5],objectPar=frame[6],font=frame[7],segPar=frame[9],mode=frame[10],postPar=frame[11]) #print('##'*20,'line64:',p_result[2]) p_result[1] = drawAllBox(p_result[2],p_result[1],frame[4],frame[5],frame[7]) time11 = time.time() image_array = p_result[1] cv2.imwrite(os.path.join('images/results/',frame[8] ) ,image_array) bname = frame[8].split('.')[0] if frame[2]: if len(p_result)==5: image_mask = p_result[4] if isinstance(image_mask,np.ndarray) and image_mask.shape[0]>0: cv2.imwrite(os.path.join('images/results/',bname+'_mask.png' ) , (image_mask).astype(np.uint8)) boxes=p_result[2] with open( os.path.join('images/results/',bname+'.txt' ),'w' ) as fp: for box in boxes: box_str=[str(x) for x in box] out_str=','.join(box_str)+'\n' fp.write(out_str) time22 = time.time() print('%s,%d*%d,AI-process: %.1f,image save:%.1f , %s'%(frame[8],H,W, (time11 - time00) * 1000.0, (time22-time11)*1000.0,timeOut)) return 'success' #except Exception as e: # return 'failed:'+str(e) def process_video(video,par0,mode='detSeg'): cap=cv2.VideoCapture(video) if not cap.isOpened(): print('#####error url:',video) return False bname=os.path.basename(video).split('.')[0] fps = int(cap.get(cv2.CAP_PROP_FPS)+0.5) width = int(cap.get(cv2.CAP_PROP_FRAME_WIDTH )+0.5) height = int(cap.get(cv2.CAP_PROP_FRAME_HEIGHT)+0.5) framecnt=int(cap.get(7)+0.5) save_path_AI = os.path.join(par0['outpth'],os.path.basename(video)) problem_image_dir= os.path.join( par0['outpth'], 'probleImages' ) os.makedirs(problem_image_dir,exist_ok=True) vid_writer_AI = cv2.VideoWriter(save_path_AI, cv2.VideoWriter_fourcc(*'mp4v'), fps, (width,height)) num=0 iframe=0;post_results=[];fpsample=30*10 imgarray_list = []; iframe_list = [] #patch_cnt = par0['trackPar']['patchCnt'] ##windowsize 对逐帧插值后的结果做平滑，windowsize为平滑的长度,没隔det_cnt帧做一次跟踪。 trackPar={'det_cnt':10,'windowsize':29 } ##track_det_result_update= np.empty((0,8)) ###每100帧跑出来的结果，放在track_det_result_update，只保留当前100帧里有的tracker Id. while cap.isOpened(): ret, imgarray = cap.read() #读取摄像头画面 iframe +=1 if not ret:break if mode=='detSeg': p_result,timeOut = AI_process([imgarray],par0['model'],par0['segmodel'],par0['names'],par0['label_arraylist'],par0['rainbows'],objectPar=par0['objectPar'],font=par0['digitFont'],segPar=par0['segPar']) else: p_result,timeOut = AI_process_forest([imgarray],par0['model'],par0['segmodel'],par0['names'],par0['label_arraylist'],par0['rainbows'],par0['half'],par0['device'],par0['conf_thres'], par0['iou_thres'],par0['allowedList'],font=par0['digitFont'],trtFlag_det=par0['trtFlag_det']) p_result[1] = drawAllBox(p_result[2],p_result[1],par0['label_arraylist'],par0['rainbows'],par0['digitFont']) if mode != 'track': image_array = p_result[1];num+=1 ret = vid_writer_AI.write(image_array) view_bar(num, framecnt,time.time(),prefix=os.path.basename(video)) ##每隔 fpsample帧处理一次，如果有问题就保存图片 if (iframe % fpsample == 0) and (len(post_results)>0) : parImage=save_problem_images(post_results,iframe,par0['names'],streamName=bname,outImaDir=problem_image_dir,imageTxtFile=False) post_results=[] if len(p_result[2] )>0: post_results.append(p_result) vid_writer_AI.release(); def detSeg_demo(opt): if opt['business'] == 'river': par={ 'device':'0', ###显卡号,如果用TRT模型，只支持0（单显卡） 'labelnames':"../AIlib2/weights/conf/river/labelnames.json", ###检测类别对照表 'max_workers':1, ###并行线程数 'Detweights':"../weights/%s/AIlib2/%s/yolov5_%s_fp16.engine"%(opt['gpu'], opt['business'] ,opt['gpu'] ),###检测模型路径 #'Detweights':"/mnt/thsw2/DSP2/weights/river/yolov5.pt",###检测模型路径 'detModelpara':[{"id":str(x),"config":{"k1":"v1","k2":"v2"}} for x in [0,1,2,3,4,5,6,7,8,9] ],###控制哪些检测类别显示，输出 'seg_nclass':2,###分割模型类别数目，默认2类 'segRegionCnt':1,###分割模型结果需要保留的等值线数目 'segPar':{'modelSize':(640,360),'mean':(0.485, 0.456, 0.406),'std' :(0.229, 0.224, 0.225),'numpy':False, 'RGB_convert_first':True,#分割模型预处理参数 'mixFunction':{'function':riverDetSegMixProcess,'pars':{'slopeIndex':[1,3,4,7], 'riverIou':0.1}} #分割和检测混合处理的函数 }, 'Segweights' : "../weights/%s/AIlib2/%s/stdc_360X640_%s_fp16.engine"%(opt['gpu'], opt['business'] ,opt['gpu'] ),###分割模型权重位置 'postFile': '../AIlib2/weights/conf/river/para.json',###后处理参数文件 'txtFontSize':40,###文本字符的大小 'digitFont': { 'line_thickness':2,'boxLine_thickness':1, 'fontSize':1.0,'waterLineColor':(0,255,255),'segLineShow':True,'waterLineWidth':3},###显示框、线、数字设置 'testImgPath':'/mnt/thsw2/DSP2/videos/river/', 'testOutPath':'images/results/',###输出测试图像位置 } if opt['business'] == 'river2': par={ 'device':'0', ###显卡号,如果用TRT模型，只支持0（单显卡） 'labelnames':"../AIlib2/weights/conf/river2/labelnames.json", ###检测类别对照表 'max_workers':1, ###并行线程数 'Detweights':"../weights/%s/AIlib2/%s/yolov5_%s_fp16.engine"%(opt['gpu'], opt['business'] ,opt['gpu'] ),###检测模型路径 'detModelpara':[{"id":str(x),"config":{"k1":"v1","k2":"v2"}} for x in [0,1,2,3,4,5,6,7,8,9] ],###控制哪些检测类别显示，输出 'seg_nclass':2,###分割模型类别数目，默认2类 'segRegionCnt':1,###分割模型结果需要保留的等值线数目 'segPar':{'modelSize':(640,360),'mean':(0.485, 0.456, 0.406),'std' :(0.229, 0.224, 0.225),'numpy':False, 'RGB_convert_first':True,#分割模型预处理参数 'mixFunction':{'function':riverDetSegMixProcess,'pars':{'slopeIndex':[1,3,4,7], 'riverIou':0.1}} #分割和检测混合处理的函数 }, 'Segweights' : "../weights/%s/AIlib2/%s/stdc_360X640_%s_fp16.engine"%(opt['gpu'], opt['business'] ,opt['gpu'] ),###分割模型权重位置 'postFile': '../AIlib2/weights/conf/river2/para.json',###后处理参数文件 'txtFontSize':40,###文本字符的大小 'digitFont': { 'line_thickness':2,'boxLine_thickness':1, 'fontSize':1.0,'waterLineColor':(0,255,255),'segLineShow':True,'waterLineWidth':3},###显示框、线、数字设置 'testImgPath':'images/river2/', 'testOutPath':'images/results/',###输出测试图像位置 } if opt['business'] == 'highWay2': par={ 'device':'0', ###显卡号,如果用TRT模型，只支持0（单显卡） 'labelnames':"../AIlib2/weights/conf/%s/labelnames.json"%(opt['business']), ###检测类别对照表 'max_workers':1, ###并行线程数 'Detweights':"../weights/%s/AIlib2/%s/yolov5_%s_fp16.engine"%(opt['gpu'], opt['business'] ,opt['gpu'] ),###检测模型路径 'detModelpara':[{"id":str(x),"config":{"k1":"v1","k2":"v2"}} for x in [0,1,2,3,5,6,7,8,9] ],###控制哪些检测类别显示、输出 'seg_nclass':3,###分割模型类别数目，默认2类 'segRegionCnt':2,###分割模型结果需要保留的等值线数目 'segPar':{'modelSize':(1920,1080),'mean':(0.485, 0.456, 0.406),'std' :(0.229, 0.224, 0.225),'predResize':True,'numpy':False, 'RGB_convert_first':True,###分割模型预处理参数 'mixFunction':{'function':tracfficAccidentMixFunction, 'pars':{ 'RoadArea': 16000, 'roadVehicleAngle': 15, 'speedRoadVehicleAngleMax': 75, 'roundness': 1.0, 'cls': 9, 'vehicleFactor': 0.1, 'confThres':0.25,'roadIou':0.6,'radius': 50 ,'vehicleFlag':False,'distanceFlag': False} } }, #'Segweights' : "../weights/%s/AIlib2/%s/stdc_360X640_%s_fp16.engine"%(opt['gpu'], opt['business'] ,opt['gpu'] ),###分割模型权重位置 'Segweights' : "../weights/%s/AIlib2/%s/stdc_1080X1920_%s_fp16.engine"%(opt['gpu'], opt['business'] ,opt['gpu'] ),###分割模型权重位置 #'Segweights' :'/mnt/thsw2/DSP2/weights/highWay2/stdc_360X640.pth', 'postFile': '../AIlib2/weights/conf/%s/para.json'%(opt['business'] ),###后处理参数文件 'txtFontSize':20,###文本字符的大小 'digitFont': { 'line_thickness':2,'boxLine_thickness':1, 'fontSize':1.0,'waterLineColor':(0,255,255),'segLineShow':True,'waterLineWidth':2},###显示框、线设置 'testImgPath':'images/highWayTest/',###测试图像的位置 #'testImgPath':'images/tt', 'testOutPath':'images/results/',###输出测试图像位置 } par['segPar']['mixFunction']['pars']['modelSize'] = par['segPar']['modelSize'] if opt['business'] == 'drowning': par={ 'device':'0', ###显卡号,如果用TRT模型，只支持0（单显卡） 'labelnames':"../AIlib2/weights/conf/%s/labelnames.json"%(opt['business']), ###检测类别对照表 'max_workers':1, ###并行线程数 'Detweights':"../weights/%s/AIlib2/%s/yolov5_%s_fp16.engine"%(opt['gpu'], opt['business'] ,opt['gpu'] ),###检测模型路径 #'Detweights':"/mnt/thsw2/DSP2/weights/drowning/yolov5.pt", 'detModelpara':[{"id":str(x),"config":{"k1":"v1","k2":"v2"}} for x in [0,1,2,3,5,6,7,8,9] ],###控制哪些检测类别显示、输出 'seg_nclass':2,###分割模型类别数目，默认2类 'segRegionCnt':2,###分割模型结果需要保留的等值线数目 'segPar':{'modelSize':(640,360),'mean':(0.485, 0.456, 0.406),'std' :(0.229, 0.224, 0.225),'predResize':True,'numpy':False, 'RGB_convert_first':True,###分割模型预处理参数 'mixFunction':{'function':mixDrowing_water_postprocess, 'pars':{ } } }, 'Segweights' : "../weights/%s/AIlib2/%s/stdc_360X640_%s_fp16.engine"%(opt['gpu'], opt['business'] ,opt['gpu'] ),###分割模型权重位置 #'Segweights' : "/mnt/thsw2/DSP2/weights/drowning/stdc_360X640_2080Ti_fp16.engine", 'postFile': '../AIlib2/weights/conf/%s/para.json'%(opt['business'] ),###后处理参数文件 'txtFontSize':20,###文本字符的大小 'digitFont': { 'line_thickness':2,'boxLine_thickness':1, 'fontSize':1.0,'waterLineColor':(0,255,255),'segLineShow':True,'waterLineWidth':2},###显示框、线设置 #'testImgPath':'/mnt/thsw2/DSP2/videos/drowning/',###测试图像的位置 'testImgPath':'images/drowning/', 'testOutPath':'images/results/',###输出测试图像位置 } par['segPar']['mixFunction']['pars']['modelSize'] = par['segPar']['modelSize'] if opt['business'] == 'noParking': par={ 'device':'0', ###显卡号,如果用TRT模型，只支持0（单显卡） 'labelnames':"../AIlib2/weights/conf/%s/labelnames.json"%(opt['business']), ###检测类别对照表 'max_workers':1, ###并行线程数 'Detweights':"../weights/%s/AIlib2/%s/yolov5_%s_fp16.engine"%(opt['gpu'], opt['business'] ,opt['gpu'] ),###检测模型路径 'detModelpara':[{"id":str(x),"config":{"k1":"v1","k2":"v2"}} for x in [0,1,2,3,5,6,7,8,9] ],###控制哪些检测类别显示、输出 'seg_nclass':4,###分割模型类别数目，默认2类 'segRegionCnt':2,###分割模型结果需要保留的等值线数目 'segPar':{'modelSize':(640,360),'mean':(0.485, 0.456, 0.406),'std' :(0.229, 0.224, 0.225),'predResize':True,'numpy':False, 'RGB_convert_first':True,###分割模型预处理参数 'mixFunction':{'function':mixNoParking_road_postprocess, 'pars':{ 'roundness': 0.3, 'cls': 9, 'laneArea': 10, 'laneAngleCha': 5 ,'RoadArea': 16000,'fitOrder':2} } }, 'Segweights' : "../weights/%s/AIlib2/%s/stdc_360X640_%s_fp16.engine"%(opt['gpu'], opt['business'] ,opt['gpu'] ),###分割模型权重位置 'postFile': '../AIlib2/weights/conf/%s/para.json'%(opt['business'] ),###后处理参数文件 'txtFontSize':20,###文本字符的大小 'digitFont': { 'line_thickness':2,'boxLine_thickness':1, 'fontSize':1.0,'waterLineColor':(0,255,255),'segLineShow':True,'waterLineWidth':2},###显示框、线设置 'testImgPath':'images/noParking/',###测试图像的位置 'testOutPath':'images/results/',###输出测试图像位置 } par['segPar']['mixFunction']['pars']['modelSize'] = par['segPar']['modelSize'] if opt['business'] == 'illParking': from utilsK.illParkingUtils import illParking_postprocess par={ 'device':'0', ###显卡号,如果用TRT模型，只支持0（单显卡） 'labelnames':"../AIlib2/weights/conf/%s/labelnames.json"%(opt['business']), ###检测类别对照表 'max_workers':1, ###并行线程数 'Detweights':"../weights/%s/AIlib2/%s/yolov5_%s_fp16.engine"%(opt['gpu'], opt['business'] ,opt['gpu'] ),###检测模型路径 'detModelpara':[{"id":str(x),"config":{"k1":"v1","k2":"v2"}} for x in [0,1,2,3,5,6,7,8,9] ],###控制哪些检测类别显示、输出 'seg_nclass':4,###分割模型类别数目，默认2类 'segRegionCnt':2,###分割模型结果需要保留的等值线数目 'segPar':{ 'mixFunction':{'function':illParking_postprocess, 'pars':{ } } }, 'Segweights' : None,###分割模型权重位置 'postFile': '../AIlib2/weights/conf/%s/para.json'%(opt['business'] ),###后处理参数文件 'txtFontSize':20,###文本字符的大小 'digitFont': { 'line_thickness':2,'boxLine_thickness':1, 'fontSize':1.0,'waterLineColor':(0,255,255),'segLineShow':True,'waterLineWidth':2},###显示框、线设置 'testImgPath':'images/cityMangement',###测试图像的位置 'testOutPath':'images/results/',###输出测试图像位置 } if opt['business'] == 'cityMangement2': from DMPR import DMPRModel from DMPRUtils.jointUtil import dmpr_yolo par={ 'device':'0', ###显卡号,如果用TRT模型，只支持0（单显卡） 'labelnames':"../AIlib2/weights/conf/%s/labelnames.json"%(opt['business']), ###检测类别对照表 'max_workers':1, ###并行线程数 'Detweights':"../weights/%s/AIlib2/%s/yolov5_%s_fp16.engine"%(opt['gpu'], opt['business'] ,opt['gpu'] ),###检测模型路径 #'Detweights':"../AIlib2/weights/conf/cityMangement2/yolov5.pt",###检测模型路径 #'Detweights':"/mnt/thsw2/DSP2/weights/cityMangement2/weights/urbanManagement/yolo/best.pt", 'detModelpara':[{"id":str(x),"config":{"k1":"v1","k2":"v2"}} for x in [0,1,2,3,5,6,7,8,9] ],###控制哪些检测类别显示、输出 'seg_nclass':4,###分割模型类别数目，默认2类 'segRegionCnt':2,###分割模型结果需要保留的等值线数目 'segPar':{ 'depth_factor':32,'NUM_FEATURE_MAP_CHANNEL':6,'dmpr_thresh':0.3, 'dmprimg_size':640, 'mixFunction':{'function':dmpr_yolo, 'pars':{'carCls':0 ,'illCls':3,'scaleRatio':0.5,'border':80} } }, #'Segweights' : '/mnt/thsw2/DSP2/weights/cityMangement2/weights/urbanManagement/DMPR/dp_detector_499.engine',###分割模型权重位置 'Segweights':"../weights/%s/AIlib2/%s/dmpr_%s.engine"%(opt['gpu'], opt['business'] ,opt['gpu'] ),###检测模型路径 #'Segweights':"../AIlib2/weights/conf/cityMangement2/dmpr.pth",###检测模型路径 'postFile': '../AIlib2/weights/conf/%s/para.json'%(opt['business'] ),###后处理参数文件 'txtFontSize':20,###文本字符的大小 'digitFont': { 'line_thickness':2,'boxLine_thickness':1, 'fontSize':1.0,'waterLineColor':(0,255,255),'segLineShow':True,'waterLineWidth':2},###显示框、线设置 #'testImgPath':'/mnt/thsw2/DSP2/demoImages/illParking',###测试图像的位置 #'testImgPath':'/mnt/thsw2/DSP2/weights/cityMangement2_0916/images/input', 'testImgPath':'images/cityMangement2/', 'testOutPath':'images/results/',###输出测试图像位置 } if par['Segweights']: par['trtFlag_seg']=True if par['Segweights'].endswith('.engine') else False else: par['trtFlag_seg']=False par['trtFlag_det']=True if par['Detweights'].endswith('.engine') else False mode = par['mode'] if 'mode' in par.keys() else 'others' postPar = par['postPar'] if 'postPar' in par.keys() else None device_=par['device'] labelnames = par['labelnames'] ##对应类别表 max_workers=par['max_workers']; trtFlag_det=par['trtFlag_det'];trtFlag_seg=par['trtFlag_seg'];segRegionCnt=par['segRegionCnt'] device = select_device(device_) names=get_labelnames(labelnames) half = device.type != 'cpu' # half precision only supported on CUDA if trtFlag_det: Detweights = par['Detweights']##升级后的检测模型 logger = trt.Logger(trt.Logger.ERROR) with open(Detweights, "rb") as f, trt.Runtime(logger) as runtime: model=runtime.deserialize_cuda_engine(f.read())# 输入trt本地文件，返回ICudaEngine对象 print('############locad det model trtsuccess:',Detweights) else: Detweights = par['Detweights'] model = attempt_load(Detweights, map_location=device) # load FP32 model print('############locad det model pth success:',Detweights) if half: model.half() par['segPar']['seg_nclass'] = par['seg_nclass'] segPar=par['segPar'] if par['Segweights']: if opt['business'] == 'cityMangement2': segmodel = DMPRModel(weights=par['Segweights'], par = par['segPar']) else: segmodel = stdcModel(weights=par['Segweights'], par = par['segPar']) else: segmodel= None print('############None seg model is loaded###########:' ) postFile= par['postFile'] digitFont= par['digitFont'] #conf_thres,iou_thres,classes,rainbows=get_postProcess_para(postFile) detPostPar = get_postProcess_para_dic(postFile) conf_thres,iou_thres,classes,rainbows = detPostPar["conf_thres"],detPostPar["iou_thres"],detPostPar["classes"],detPostPar["rainbows"] if 'ovlap_thres_crossCategory' in detPostPar.keys(): ovlap_thres_crossCategory=detPostPar['ovlap_thres_crossCategory'] else:ovlap_thres_crossCategory = None if 'score_byClass' in detPostPar.keys(): score_byClass=detPostPar['score_byClass'] else: score_byClass = None ####模型选择参数用如下： mode_paras=par['detModelpara'] allowedList,allowedList_string=get_needed_objectsIndex(mode_paras) #allowedList=[0,1,2,3] ##加载模型，准备好显示字符 label_arraylist = get_label_arrays(names,rainbows,outfontsize=par['txtFontSize'],fontpath="../AIlib2/conf/platech.ttf") ##图像测试 #impth = 'images/slope/' impth = par['testImgPath'] outpth = par['testOutPath'] imgpaths=[]###获取文件里所有的图像 for postfix in ['.jpg','.JPG','.PNG','.png']: imgpaths.extend(glob.glob('%s/*%s'%(impth,postfix )) ) videopaths=[]###获取文件里所有的视频 for postfix in ['.MP4','.mp4','.avi']: videopaths.extend(glob.glob('%s/*%s'%(impth,postfix )) ) ###先处理图像 frames=[] for imgpath in imgpaths: im0s=[cv2.imread(imgpath)] objectPar={ 'half':half,'device':device,'conf_thres':conf_thres,'ovlap_thres_crossCategory':ovlap_thres_crossCategory,'iou_thres':iou_thres,'allowedList':allowedList,'segRegionCnt':segRegionCnt, 'trtFlag_det':trtFlag_det,'trtFlag_seg':trtFlag_seg ,'score_byClass':score_byClass} #p_result[1] = draw_painting_joint(xyxy,p_result[1],label_arraylist[int(cls)],score=conf,color=rainbows[int(cls)%20],font=font,socre_location="leftBottom") frame=(im0s,model,segmodel,names,label_arraylist,rainbows,objectPar,digitFont,os.path.basename(imgpath),segPar,mode,postPar) frames.append(frame) t1=time.time() if max_workers==1: for i in range(len(imgpaths)): print('-'*20,imgpaths[i],'-'*20) t5=time.time() process_v1(frames[i]) t6=time.time() #print('#######%s, ms:%.1f , accumetate time:%.1f, avage:%1.f '%(os.path.basename(imgpaths[i]), (t6-t5)*1000.0,(t6-t1)*1000.0, (t6-t1)*1000.0/(i+1))) else: with ThreadPoolExecutor(max_workers=max_workers) as t: for result in t.map(process_v1, frames): #print(result) t=result t2=time.time() if len(imgpaths)>0: print('All %d images time:%.1f ms ,each:%.1f ms, with %d threads'%(len(imgpaths),(t2-t1)*1000, (t2-t1)*1000.0/len(imgpaths) , max_workers) ) objectPar={ 'half':half,'device':device,'conf_thres':conf_thres,'iou_thres':iou_thres,'allowedList':allowedList,'segRegionCnt':segRegionCnt, 'trtFlag_det':trtFlag_det,'trtFlag_seg':trtFlag_seg } par0={ 'model':model,'segmodel':segmodel,'names':names,'label_arraylist':label_arraylist,'rainbows':rainbows, 'objectPar':objectPar,'digitFont':digitFont,'segPar':segPar,'outpth':outpth } ###如果是视频文件 for video in videopaths: process_video(video,par0) print(' ') def det_demo(business ): ####森林巡检的参数 if opt['business'] == 'forest': par={ 'device':'0', ###显卡号,如果用TRT模型，只支持0（单显卡） 'labelnames':"../AIlib2/weights/conf/forest/labelnames.json", ###检测类别对照表 'gpuname':'3090',###显卡名称 'max_workers':1, ###并行线程数 'trtFlag_det':True,###检测模型是否采用TRT 'trtFlag_seg':False,###分割模型是否采用TRT 'Detweights':"../weights/%s/AIlib2/%s/yolov5_%s_fp16.engine"%(opt['gpu'], opt['business'] ,opt['gpu'] ),###检测模型路径 'detModelpara':[{"id":str(x),"config":{"k1":"v1","k2":"v2"}} for x in [0,1,2,3,4,5,6] ],###控制哪些检测类别显示、输出 'slopeIndex':[],###岸坡类别（或者其它业务里的类别），不与河道（分割的前景区域）计算交并比，即不论是否在河道内都显示。 'seg_nclass':2,###分割模型类别数目，默认2类 'segRegionCnt':0,###分割模型结果需要保留的等值线数目 'segPar':None,###分割模型预处理参数 'Segweights' : None,###分割模型权重位置 'postFile': '../AIlib2/weights/conf/forest/para.json',###后处理参数文件 'txtFontSize':80,###文本字符的大小 'digitFont': { 'line_thickness':2,'boxLine_thickness':1, 'fontSize':1.0,'waterLineColor':(0,255,255),'waterLineWidth':3},###显示框、线设置 'testImgPath':'../AIdemo2/images/forest/',###测试图像的位置 'testOutPath':'images/results/',###输出测试图像位置 } # if opt['business'] == 'forest2': par={ 'device':'0', ###显卡号,如果用TRT模型，只支持0（单显卡） 'labelnames':"../AIlib2/weights/conf/forest2/labelnames.json", ###检测类别对照表 'gpuname':opt['gpu'],###显卡名称 'max_workers':1, ###并行线程数 'trtFlag_det':True,###检测模型是否采用TRT 'trtFlag_seg':False,###分割模型是否采用TRT 'Detweights':"../weights/%s/AIlib2/%s/yolov5_%s_fp16.engine"%(opt['gpu'], opt['business'] ,opt['gpu'] ),###检测模型路径 'detModelpara':[{"id":str(x),"config":{"k1":"v1","k2":"v2"}} for x in [0,1,2,3,4,5,6] ],###控制哪些检测类别显示、输出 #'detModelpara':[{"id":str(x),"config":{"k1":"v1","k2":"v2"}} for x in [] ], 'slopeIndex':[],###岸坡类别（或者其它业务里的类别），不与河道（分割的前景区域）计算交并比，即不论是否在河道内都显示。 'seg_nclass':2,###分割模型类别数目，默认2类 'segRegionCnt':0,###分割模型结果需要保留的等值线数目 'segPar':None,###分割模型预处理参数 'Segweights' : None,###分割模型权重位置 'postFile': '../AIlib2/weights/conf/forest/para.json',###后处理参数文件 'txtFontSize':80,###文本字符的大小 'digitFont': { 'line_thickness':2,'boxLine_thickness':1, 'fontSize':1.0,'waterLineColor':(0,255,255),'waterLineWidth':3},###显示框、线设置 'testImgPath':'../AIdemo2/images/forest2/',###测试图像的位置 'testOutPath':'images/results/',###输出测试图像位置 } ###车辆巡检参数 if opt['business'] == 'vehicle': par={ 'device':'0', ###显卡号,如果用TRT模型，只支持0（单显卡） 'labelnames':"../AIlib2/weights/conf/vehicle/labelnames.json", ###检测类别对照表 'gpuname':'2080T',###显卡名称 'max_workers':1, ###并行线程数 'trtFlag_det':True,###检测模型是否采用TRT 'trtFlag_seg':False,###分割模型是否采用TRT 'Detweights':"../weights/%s/AIlib2/%s/yolov5_%s_fp16.engine"%(opt['gpu'], opt['business'] ,opt['gpu'] ),###检测模型路径 'detModelpara':[{"id":str(x),"config":{"k1":"v1","k2":"v2"}} for x in [0,1,2,3,4,5,6] ],###控制哪些检测类别显示、输出 'slopeIndex':[],###岸坡类别（或者其它业务里的类别），不与河道（分割的前景区域）计算交并比，即不论是否在河道内都显示。 'seg_nclass':2,###分割模型类别数目，默认2类 'segRegionCnt':0,###分割模型结果需要保留的等值线数目 'segPar':None,###分割模型预处理参数 'Segweights' : None,###分割模型权重位置 'postFile': '../AIlib2/weights/conf/vehicle/para.json',###后处理参数文件 'txtFontSize':40,###文本字符的大小 'digitFont': { 'line_thickness':2,'boxLine_thickness':1, 'fontSize':1.0,'waterLineColor':(0,255,255),'waterLineWidth':3},###显示框、线设置 'testImgPath':'../AIdemo2/images/vehicle/',###测试图像的位置 'testOutPath':'images/results/',###输出测试图像位置 } ###行人检测模型 if opt['business'] == 'pedestrian': par={ 'device':'0', ###显卡号,如果用TRT模型，只支持0（单显卡） 'labelnames':"../AIlib2/weights/conf/pedestrian/labelnames.json", ###检测类别对照表 'gpuname':'2080T',###显卡名称 'max_workers':1, ###并行线程数 'trtFlag_det':True,###检测模型是否采用TRT 'trtFlag_seg':False,###分割模型是否采用TRT 'Detweights':"../weights/%s/AIlib2/%s/yolov5_%s_fp16.engine"%(opt['gpu'], opt['business'] ,opt['gpu'] ),###检测模型路径 'detModelpara':[{"id":str(x),"config":{"k1":"v1","k2":"v2"}} for x in [0,1,2,3,4,5,6] ],###控制哪些检测类别显示、输出 'slopeIndex':[],###岸坡类别（或者其它业务里的类别），不与河道（分割的前景区域）计算交并比，即不论是否在河道内都显示。 'seg_nclass':2,###分割模型类别数目，默认2类 'segRegionCnt':0,###分割模型结果需要保留的等值线数目 'segPar':None,###分割模型预处理参数 'Segweights' : None,###分割模型权重位置 'postFile': '../AIlib2/weights/conf/pedestrian/para.json',###后处理参数文件 'txtFontSize':40,###文本字符的大小 'digitFont': { 'line_thickness':2,'boxLine_thickness':1, 'fontSize':1.0,'waterLineColor':(0,255,255),'waterLineWidth':3},###显示框、线设置 'testImgPath':'../AIdemo2/images/pedestrian/',###测试图像的位置 'testOutPath':'images/results/',###输出测试图像位置 } ###烟雾火焰检测模型 if opt['business'] == 'smogfire': par={ 'device':'0', ###显卡号,如果用TRT模型，只支持0（单显卡） 'labelnames':"../AIlib2/weights/conf/smogfire/labelnames.json", ###检测类别对照表 'gpuname':'2080T',###显卡名称 'max_workers':1, ###并行线程数 'trtFlag_det':True,###检测模型是否采用TRT 'trtFlag_seg':False,###分割模型是否采用TRT 'Detweights':"../weights/%s/AIlib2/%s/yolov5_%s_fp16.engine"%(opt['gpu'], opt['business'] ,opt['gpu'] ),###检测模型路径 'detModelpara':[{"id":str(x),"config":{"k1":"v1","k2":"v2"}} for x in [0,1,2,3,4,5,6] ],###控制哪些检测类别显示、输出 'slopeIndex':[],###岸坡类别（或者其它业务里的类别），不与河道（分割的前景区域）计算交并比，即不论是否在河道内都显示。 'seg_nclass':2,###没有分割模型，此处不用 'segRegionCnt':0,###没有分割模型，此处不用 'segPar':None,###分割模型预处理参数 'Segweights' : None,###分割模型权重位置 'postFile': '../AIlib2/weights/conf/smogfire/para.json',###后处理参数文件 'txtFontSize':40,###文本字符的大小 'digitFont': { 'line_thickness':2,'boxLine_thickness':1, 'fontSize':1.0,'waterLineColor':(0,255,255),'waterLineWidth':3},###显示框、线设置 'testImgPath':'../AIdemo2/images/smogfire/',###测试图像的位置 'testOutPath':'images/results/',###输出测试图像位置 } ###钓鱼游泳检测 if opt['business'] == 'AnglerSwimmer': par={ 'device':'0', ###显卡号,如果用TRT模型，只支持0（单显卡） 'labelnames':"../AIlib2/weights/conf/AnglerSwimmer/labelnames.json", ###检测类别对照表 'gpuname':'2080T',###显卡名称 'max_workers':1, ###并行线程数 'trtFlag_det':True,###检测模型是否采用TRT 'trtFlag_seg':False,###分割模型是否采用TRT 'Detweights':"../weights/%s/AIlib2/%s/yolov5_%s_fp16.engine"%(opt['gpu'], opt['business'] ,opt['gpu'] ),###检测模型路径 'detModelpara':[{"id":str(x),"config":{"k1":"v1","k2":"v2"}} for x in [0,1,2,3,4,5,6] ],###控制哪些检测类别显示、输出 'slopeIndex':[],###岸坡类别（或者其它业务里的类别），不与河道（分割的前景区域）计算交并比，即不论是否在河道内都显示。 'seg_nclass':2,###没有分割模型，此处不用 'segRegionCnt':0,###没有分割模型，此处不用 'segPar':None,###分割模型预处理参数 'Segweights' : None,###分割模型权重位置 'postFile': '../AIlib2/weights/conf/AnglerSwimmer/para.json',###后处理参数文件 'txtFontSize':40,###文本字符的大小 'digitFont': { 'line_thickness':2,'boxLine_thickness':1, 'fontSize':1.0,'waterLineColor':(0,255,255),'waterLineWidth':3},###显示框、线设置 'testImgPath':'../AIdemo2/images/AnglerSwimmer/',###测试图像的位置 'testOutPath':'images/results/',###输出测试图像位置 } ###航道应急，做落水人员检测, channelEmergency if opt['business'] == 'channelEmergency': par={ 'device':'0', ###显卡号,如果用TRT模型，只支持0（单显卡） 'labelnames':"../AIlib2/weights/conf/channelEmergency/labelnames.json", ###检测类别对照表 'gpuname':'2080T',###显卡名称 'max_workers':1, ###并行线程数 'trtFlag_det':True,###检测模型是否采用TRT 'trtFlag_seg':False,###分割模型是否采用TRT 'Detweights':"../weights/%s/AIlib2/%s/yolov5_%s_fp16.engine"%(opt['gpu'], opt['business'] ,opt['gpu'] ),###检测模型路径 #'detModelpara':[{"id":str(x),"config":{"k1":"v1","k2":"v2"}} for x in [0,1,2,3,4,5,6] ],###控制哪些检测类别显示、输出 'detModelpara':[{"id":str(x),"config":{"k1":"v1","k2":"v2"}} for x in [] ],###控制哪些检测类别显示、输出 'slopeIndex':[],###岸坡类别（或者其它业务里的类别），不与河道（分割的前景区域）计算交并比，即不论是否在河道内都显示。 'seg_nclass':2,###没有分割模型，此处不用 'segRegionCnt':0,###没有分割模型，此处不用 'segPar':None,###分割模型预处理参数 'Segweights' : None,###分割模型权重位置 'postFile': '../AIlib2/weights/conf/channelEmergency/para.json',###后处理参数文件 'txtFontSize':40,###文本字符的大小 'digitFont': { 'line_thickness':2,'boxLine_thickness':1, 'fontSize':1.0,'waterLineColor':(0,255,255),'waterLineWidth':3},###显示框、线设置 'testImgPath':'../AIdemo2/images/channelEmergency/',###测试图像的位置 'testOutPath':'images/results/',###输出测试图像位置 } ###乡村路违法种植 if opt['business'] == 'countryRoad': par={ 'device':'0', ###显卡号,如果用TRT模型，只支持0（单显卡） 'labelnames':"../AIlib2/weights/conf/countryRoad/labelnames.json", ###检测类别对照表 'gpuname':'2080T',###显卡名称 'max_workers':1, ###并行线程数 'trtFlag_det':True,###检测模型是否采用TRT 'trtFlag_seg':False,###分割模型是否采用TRT 'Detweights':"../weights/%s/AIlib2/%s/yolov5_%s_fp16.engine"%(opt['gpu'], opt['business'] ,opt['gpu'] ),###检测模型路径 'detModelpara':[{"id":str(x),"config":{"k1":"v1","k2":"v2"}} for x in [0,1,2,3,4,5,6] ],###控制哪些检测类别显示、输出 'slopeIndex':[],###岸坡类别（或者其它业务里的类别），不与河道（分割的前景区域）计算交并比，即不论是否在河道内都显示。 'seg_nclass':2,###没有分割模型，此处不用 'segRegionCnt':0,###没有分割模型，此处不用 'segPar':None,###分割模型预处理参数 'Segweights' : None,###分割模型权重位置 'postFile': '../AIlib2/weights/conf/countryRoad/para.json',###后处理参数文件 'txtFontSize':40,###文本字符的大小 'digitFont': { 'line_thickness':2,'boxLine_thickness':1, 'fontSize':1.0,'waterLineColor':(0,255,255),'waterLineWidth':3},###显示框、线设置 'testImgPath':'../AIdemo2/images/countryRoad/',###测试图像的位置 'testOutPath':'images/results/',###输出测试图像位置 } ###河道上大型船只 if opt['business'] == 'ship': par={ 'device':'0', ###显卡号,如果用TRT模型，只支持0（单显卡） 'labelnames':"../AIlib2/weights/conf/%s/labelnames.json"%(opt['business']), ###检测类别对照表 'gpuname':'2080T',###显卡名称 'max_workers':1, ###并行线程数 'trtFlag_det':True,###检测模型是否采用TRT 'trtFlag_seg':False,###分割模型是否采用TRT 'Detweights':"../weights/%s/AIlib2/%s/yolov5_%s_fp16.engine"%(opt['gpu'], opt['business'] ,opt['gpu'] ),###检测模型路径 'detModelpara':[{"id":str(x),"config":{"k1":"v1","k2":"v2"}} for x in [0,1,2,3,4,5,6] ],###控制哪些检测类别显示、输出 'slopeIndex':[],###岸坡类别（或者其它业务里的类别），不与河道（分割的前景区域）计算交并比，即不论是否在河道内都显示。 'seg_nclass':2,###没有分割模型，此处不用 'segRegionCnt':0,###没有分割模型，此处不用 'segPar':None,###分割模型预处理参数 'Segweights' : None,###分割模型权重位置 'postFile': '../AIlib2/weights/conf/%s/para.json'%(opt['business']),###后处理参数文件 'txtFontSize':40,###文本字符的大小 'digitFont': { 'line_thickness':2,'boxLine_thickness':1, 'fontSize':1.0,'waterLineColor':(0,255,255),'waterLineWidth':3},###显示框、线设置 'testImgPath':'../../../data/XunHe/shipData/',###测试图像的位置 'testOutPath':'images/results/',###输出测试图像位置 } ###城管项目，检测城市垃圾和车辆 if opt['business'] == 'cityMangement': par={ 'device':'0', ###显卡号,如果用TRT模型，只支持0（单显卡） 'labelnames':"../AIlib2/weights/conf/%s/labelnames.json"%(opt['business']), ###检测类别对照表 'gpuname':'2080Ti',###显卡名称 'max_workers':1, ###并行线程数 'trtFlag_det':True,###检测模型是否采用TRT 'trtFlag_seg':False,###分割模型是否采用TRT 'Detweights':"../weights/%s/AIlib2/%s/yolov5_%s_fp16.engine"%(opt['gpu'], opt['business'] ,opt['gpu'] ),###检测模型路径 'detModelpara':[{"id":str(x),"config":{"k1":"v1","k2":"v2"}} for x in [0,1,2,3,4,5,6] ],###控制哪些检测类别显示、输出 'slopeIndex':[],###岸坡类别（或者其它业务里的类别），不与河道（分割的前景区域）计算交并比，即不论是否在河道内都显示。 'seg_nclass':2,###没有分割模型，此处不用 'segRegionCnt':0,###没有分割模型，此处不用 'segPar':None,###分割模型预处理参数 'Segweights' : None,###分割模型权重位置 'postFile': '../AIlib2/weights/conf/%s/para.json'%(opt['business']),###后处理参数文件 'txtFontSize':40,###文本字符的大小 'digitFont': { 'line_thickness':2,'boxLine_thickness':1, 'fontSize':1.0,'waterLineColor':(0,255,255),'waterLineWidth':3},###显示框、线设置 'testImgPath':'images/tmp',###测试图像的位置 'testOutPath':'images/results/',###输出测试图像位置 } ###城管项目，检测道路情况，输入类别为五个："护栏","交通标志","非交通标志","施工","施工“(第4，第5类别合并，名称相同) ###实际模型检测输出的类别为："护栏","交通标志","非交通标志","锥桶","水马" if opt['business'] == 'cityRoad': par={ 'device':'0', ###显卡号,如果用TRT模型，只支持0（单显卡） 'labelnames':"../AIlib2/weights/conf/%s/labelnames.json"%(opt['business']), ###检测类别对照表 'gpuname':'2080Ti',###显卡名称 'max_workers':1, ###并行线程数 'trtFlag_det':True,###检测模型是否采用TRT 'trtFlag_seg':False,###分割模型是否采用TRT 'Detweights':"../weights/%s/AIlib2/%s/yolov5_%s_fp16.engine"%(opt['gpu'], opt['business'] ,opt['gpu'] ),###检测模型路径 'detModelpara':[{"id":str(x),"config":{"k1":"v1","k2":"v2"}} for x in [0,1,2,3,4,5,6] ],###控制哪些检测类别显示、输出 'slopeIndex':[],###岸坡类别（或者其它业务里的类别），不与河道（分割的前景区域）计算交并比，即不论是否在河道内都显示。 'seg_nclass':2,###没有分割模型，此处不用 'segRegionCnt':0,###没有分割模型，此处不用 'segPar':None,###分割模型预处理参数 'Segweights' : None,###分割模型权重位置 'postFile': '../AIlib2/weights/conf/%s/para.json'%(opt['business']),###后处理参数文件 'txtFontSize':40,###文本字符的大小 'digitFont': { 'line_thickness':2,'boxLine_thickness':1, 'fontSize':1.0,'waterLineColor':(0,255,255),'waterLineWidth':3},###显示框、线设置 'testImgPath':'images/%s'%(opt['business'] ),###测试图像的位置 'testOutPath':'images/results/',###输出测试图像位置 } #segRegionCnt=par['segRegionCnt'] trtFlag_seg = par['trtFlag_seg'];segPar=par['segPar'] ##使用森林，道路模型,business 控制['forest','road'] ##预先设置的参数 gpuname=par['gpuname']#如果用trt就需要此参数，只能是"3090" "2080Ti" device_=par['device'] ##选定模型，可选 cpu,'0','1' device = select_device(device_) half = device.type != 'cpu' # half precision only supported on CUDA trtFlag_det=par['trtFlag_det'] ###是否采用TRT模型加速 ##以下参数目前不可改 imageW=1536 ####道路模型 digitFont= par['digitFont'] if trtFlag_det: Detweights=par['Detweights'] logger = trt.Logger(trt.Logger.ERROR) with open(Detweights, "rb") as f, trt.Runtime(logger) as runtime: model=runtime.deserialize_cuda_engine(f.read())# 输入trt本地文件，返回ICudaEngine对象 print('####load TRT model :%s'%(Detweights)) else: Detweights=par['Detweights'] model = attempt_load(Detweights, map_location=device) # load FP32 model if half: model.half() labelnames = par['labelnames'] postFile= par['postFile'] print( Detweights,labelnames ) #conf_thres,iou_thres,classes,rainbows=get_postProcess_para(postFile) detPostPar = get_postProcess_para_dic(postFile) conf_thres,iou_thres,classes,rainbows = detPostPar["conf_thres"],detPostPar["iou_thres"],detPostPar["classes"],detPostPar["rainbows"] if 'ovlap_thres_crossCategory' in detPostPar.keys(): ovlap_thres_crossCategory=detPostPar['ovlap_thres_crossCategory'] else:ovlap_thres_crossCategory = None ####模型选择参数用如下： mode_paras=par['detModelpara'] allowedList,allowedList_string=get_needed_objectsIndex(mode_paras) slopeIndex = par['slopeIndex'] ##只加载检测模型，准备好显示字符 names=get_labelnames(labelnames) #imageW=4915;###默认是1920，在森林巡检的高清图像中是4920 outfontsize=int(imageW/1920*40);### label_arraylist = get_label_arrays(names,rainbows,outfontsize=par['txtFontSize'],fontpath="../AIlib2/conf/platech.ttf") segmodel = None ##图像测试 #url='images/examples/20220624_响水河_12300_1621.jpg' impth = par['testImgPath'] outpth = par['testOutPath'] imgpaths=[]###获取文件里所有的图像 for postfix in ['.jpg','.JPG','.PNG','.png']: imgpaths.extend(glob.glob('%s/*%s'%(impth,postfix )) ) videopaths=[]###获取文件里所有的视频 for postfix in ['.MP4','.mp4','.avi']: videopaths.extend(glob.glob('%s/*%s'%(impth,postfix )) ) imgpaths.sort() for i in range(len(imgpaths)): #for i in range(2): #imgpath = os.path.join(impth, folders[i]) imgpath = imgpaths[i] bname = os.path.basename(imgpath ) im0s=[cv2.imread(imgpath)] time00 = time.time() #使用不同的函数。每一个领域采用一个函数 p_result,timeOut = AI_process_forest(im0s,model,segmodel,names,label_arraylist,rainbows,half,device,conf_thres, iou_thres,allowedList,font=digitFont,trtFlag_det=trtFlag_det,SecNms=ovlap_thres_crossCategory) time11 = time.time() image_array = p_result[1] cv2.imwrite( os.path.join( outpth,bname ) ,image_array ) print('----image:%s, process:%.1f ,save:%.1f, %s'%(bname,(time11-time00) * 1000, (time.time() - time11) * 1000,timeOut ) ) ##process video print('##begin to process videos, total %d videos'%( len(videopaths))) for i,video in enumerate(videopaths): print('process video%d :%s '%(i,video)) par0={'model':model,'segmodel':segmodel, 'names':names,'label_arraylist':label_arraylist,'rainbows':rainbows,'outpth':par['testOutPath'], 'half':half,'device':device,'conf_thres':conf_thres, 'iou_thres':iou_thres,'allowedList':allowedList,'digitFont':digitFont,'trtFlag_det': trtFlag_det } process_video(video,par0,mode='det') def OCR_demo2(opt): from ocrUtils2 import crnn_model from ocrUtils2.ocrUtils import get_cfg,recognition_ocr,strLabelConverter if opt['business'] == 'ocr2': par={ 'image_dir':'images/ocr_en', 'outtxt':'images/results', 'weights':'../AIlib2/weights/conf/ocr2/crnn_448X32.pth', #'weights':'../weights/2080Ti/AIlib2/ocr2/crnn_2080Ti_fp16_448X32.engine', 'device':'cuda:0', 'cfg':'../AIlib2/weights/conf/ocr2/360CC_config.yaml', 'char_file':'../AIlib2/weights/conf/ocr2/chars.txt', 'imgH':32, 'imgW':448, 'workers':1 } image_dir=par['image_dir'] outtxt=par['outtxt'] workers=par['workers'] weights= par['weights'] device=par['device'] char_file=par['char_file'] imgH=par['imgH'] imgW=par['imgW'] cfg = par['cfg'] config = get_cfg(cfg, char_file) par['contextFlag']=False device = torch.device('cuda:0') if torch.cuda.is_available() else torch.device('cpu') if weights.endswith('.pth'): model = crnn_model.get_crnn(config,weights=weights).to(device) par['model_mode']='pth' else: logger = trt.Logger(trt.Logger.ERROR) with open(weights, "rb") as f, trt.Runtime(logger) as runtime: model = runtime.deserialize_cuda_engine(f.read())# 输入trt本地文件，返回ICudaEngine对象 print('#####load TRT file:',weights,'success #####') context = model.create_execution_context() par['model_mode']='trt';par['contextFlag']=context converter = strLabelConverter(config.DATASET.ALPHABETS) img_urls=glob.glob('%s/*.jpg'%( image_dir )) img_urls.extend( glob.glob('%s/*.png'%( image_dir )) ) cnt=len(img_urls) print('%s has %d images'%(image_dir ,len(img_urls) ) ) # 准备数据 parList=[] for i in range(cnt): img_patch=cv2.imread( img_urls[i] , cv2.IMREAD_GRAYSCALE) started = time.time() img = cv2.imread(img_urls[i]) sim_pred = recognition_ocr(config, img, model, converter, device,par=par) finished = time.time() print('{0}: elapsed time: {1} prd:{2} '.format( os.path.basename( img_urls[i] ), finished - started, sim_pred )) def OBB_demo(opt): ###倾斜框（OBB）的ship目标检测 par={ 'model_size':(608,608), #width,height 'K':100, #Maximum of objects' 'conf_thresh':0.18,##Confidence threshold, 0.1 for general evaluation 'device':"cuda:0", 'down_ratio':4,'num_classes':15, #'weights':'../AIlib2/weights/conf/ship2/obb_608X608.engine', 'weights':'../weights/%s/AIlib2/%s/obb_608X608_%s_fp16.engine'%(opt['gpu'],opt['business'],opt['gpu']), 'dataset':'dota', 'test_dir': 'images/ship/', 'result_dir': 'images/results', 'half': False, 'mean':(0.5, 0.5, 0.5), 'std':(1, 1, 1), 'model_size':(608,608),##width,height 'heads': {'hm': None,'wh': 10,'reg': 2,'cls_theta': 1}, 'decoder':None, 'test_flag':True, 'postFile': '../AIlib2/weights/conf/%s/para.json'%(opt['business'] ),###后处理参数文件 'drawBox':False,#####是否画框 'digitWordFont': { 'line_thickness':2,'boxLine_thickness':1,'wordSize':40, 'fontSize':1.0,'label_location':'leftTop'}, 'labelnames':"../AIlib2/weights/conf/%s/labelnames.json"%(opt['business'] ), ###检测类别对照表 } ####加载模型 model,decoder2=load_model_decoder_OBB(par) par['decoder']=decoder2 names=get_labelnames(par['labelnames']);par['labelnames']=names conf_thres,iou_thres,classes,rainbows=get_postProcess_para(par['postFile']);par['rainbows']=rainbows label_arraylist = get_label_arrays(names,rainbows,outfontsize=par['digitWordFont']['wordSize'],fontpath="../AIlib2/conf/platech.ttf") par['label_array']=label_arraylist img_urls=glob.glob('%s/*'%( par['test_dir'] )) for img_url in img_urls: #print(img_url) ori_image=cv2.imread(img_url) ori_image_list,infos = OBB_infer(model,ori_image,par) ori_image_list[1] = draw_obb(ori_image_list[2] ,ori_image_list[1],par) imgName = os.path.basename(img_url) saveFile = os.path.join(par['result_dir'], imgName) ret=cv2.imwrite(saveFile, ori_image_list[1] ) if not ret: print(saveFile, ' not created ') print( os.path.basename(img_url),':',infos) def jkm_demo(): from utilsK.jkmUtils import pre_process,post_process,get_return_data img_type = 'plate' ## code,plate par={'code':{'weights':'../AIlib2/weights/jkm/health_yolov5s_v3.jit','img_type':'code','nc':10 }, 'plate':{'weights':'../AIlib2/weights/jkm/plate_yolov5s_v3.jit','img_type':'plate','nc':1 }, 'conf_thres': 0.4, 'iou_thres':0.45, 'device':'cuda:0', 'plate_dilate':(0.5,0.1) } ###加载模型 device = torch.device(par['device']) jit_weights = par['code']['weights'] model = torch.jit.load(jit_weights) jit_weights = par['plate']['weights'] model_plate = torch.jit.load(jit_weights) imgd='images/plate' imgpaths = os.listdir(imgd) for imgp in imgpaths[0:]: #imgp = 'plate_IMG_20221030_100612.jpg' imgpath = os.path.join(imgd,imgp) im0 = cv2.imread(imgpath) #读取数据 img ,padInfos = pre_process(im0,device) ##预处理 if img_type=='code': pred = model(img) ##模型推理 else: pred = model_plate(img) boxes = post_process(pred,padInfos,device,conf_thres= par['conf_thres'], iou_thres= par['iou_thres'],nc=par[img_type]['nc']) #后处理 dataBack=get_return_data(im0,boxes,modelType=img_type,plate_dilate=par['plate_dilate']) print(imgp,boxes,dataBack['type']) for key in dataBack.keys(): if isinstance(dataBack[key],list): cv2.imwrite( 'images/results/%s_%s.jpg'%( imgp.replace('.jpg','').replace('.png',''),key),dataBack[key][0] ) ###返回值： dataBack def crowd_demo(opt): if opt['business']=='crowdCounting': from crowd import crowdModel as Model par={ 'mean':[0.485, 0.456, 0.406], 'std':[0.229, 0.224, 0.225],'threshold':0.5, 'input_profile_shapes':[(1,3,256,256),(1,3,1024,1024),(1,3,2048,2048)], 'modelPar':{'backbone':'vgg16_bn', 'gpu_id':0,'anchorFlag':False, 'width':None,'height':None ,'line':2, 'row':2}, 'weights':"../weights/%s/AIlib2/%s/crowdCounting_%s_dynamic.engine"%(opt['gpu'], opt['business'] ,opt['gpu'] ),###检测模型路径 'testImgPath':'images/%s'%(opt['business'] ),###测试图像的位置 'testOutPath':'images/results/',###输出测试图像位置 } #weights='weights/best_mae.pth' cmodel = Model(par['weights'],par) img_path = par['testImgPath'] File = os.listdir(img_path) targetList = [] for file in File[0:]: COORlist = [] imgPath = img_path + os.sep + file img_raw = cv2.cvtColor(cv2.imread(imgPath),cv2.COLOR_BGR2RGB) # cmodel.eval--- # 输入读取的RGB数组 # 输出:list,0--原图，1-人头坐标list,2-对接OBB的格式数据，其中4个坐标均相同，2-格式如下： # [ [ [ (x0,y0),(x1,y1),(x2,y2),(x3,y3) ],score， cls ], [ [ (x0,y0),(x1,y1),(x2,y2),(x3,y3) ],score ，cls ],........ ] prets, infos = cmodel.eval(img_raw) print(file,infos,' 人数：',len(prets[1])) img_to_draw = cv2.cvtColor(np.array(img_raw), cv2.COLOR_RGB2BGR) # 打印预测图像中人头的个数 for p in prets[1]: img_to_draw = cv2.circle(img_to_draw, (int(p[0]), int(p[1])), 2, (0, 255, 0), -1) COORlist.append((int(p[0]), int(p[1]))) # 将各测试图像中的人头坐标存储在targetList中, 格式：[[(x1, y1),(x2, y2),...], [(X1, Y1),(X2, Y2),..], ...] targetList.append(COORlist) #time.sleep(2) # 保存预测图片 cv2.imwrite(os.path.join(par['testOutPath'], file), img_to_draw) if __name__=="__main__": #jkm_demo() businessAll=['river2','AnglerSwimmer', 'countryRoad','forest2', 'pedestrian' , 'smogfire' , 'vehicle','ship2',"highWay2","channelEmergency","cityMangement","drowning","noParking","illParking",'cityMangement2',"cityRoad","crowdCounting"] businessAll = ['crowdCounting'] # forest 、 ocr2 、ocr_en 、 river 、 road 、 ship ,目前都没有在用 for busi in businessAll: print('-'*40,'beg to test ',busi,'-'*40) opt={'gpu':'2080Ti','business':busi} if opt['business'] in ['highWay2','river2','drowning','noParking','river',"illParking","cityMangement2"]: detSeg_demo(opt) elif opt['business'] in ['crowdCounting'] : crowd_demo(opt) elif opt['business'] in ['ship2']: OBB_demo(opt) elif opt['business'] in ['ocr']: OCR_demo(opt) elif opt['business'] in ['ocr2']: OCR_demo2(opt) elif opt['business'] in ['riverTrack','highWay2Track']: det_track_demo(opt ) else: det_demo( opt )