wangjin0928
/
AIdemo2
Seguir 1
Destacar 0
Fork 0
Código Incidencias 0 Pull Requests 0 Lanzamientos 0 Wiki Actividad
No puede seleccionar más de 25 temas Los temas deben comenzar con una letra o número, pueden incluir guiones ('-') y pueden tener hasta 35 caracteres de largo.
56 Commits
1 Rama
Árbol: 738dbee1d1
Ramas Etiquetas
${ item.name }
Crear rama ${ searchTerm }
desde '738dbee1d1'
${ noResults }
AIdemo2/demo3.0.py

1141 líneas
67KB
Original Blame Histórico 

  1. import sys, yaml

  2. from easydict import EasyDict as edict

  3. from concurrent.futures import ThreadPoolExecutor

  4. 

  5. sys.path.extend(['..','../AIlib2' ])

  6. 

  7. from AI import AI_process,AI_process_forest,get_postProcess_para,get_postProcess_para_dic,ocr_process,AI_det_track,AI_det_track_batch,AI_det_track_batch_N,get_images_videos,default_mix

  8. import cv2,os,time

  9. from segutils.segmodel import SegModel

  10. from yolov5 import yolov5Model

  11. from stdc import stdcModel

  12. from segutils.trafficUtils import tracfficAccidentMixFunction

  13. from models.experimental import attempt_load

  14. from utils.torch_utils import select_device

  15. from utilsK.queRiver import get_labelnames,get_label_arrays,save_problem_images,riverDetSegMixProcess,draw_painting_joint

  16. from ocrUtils.ocrUtils import CTCLabelConverter,AlignCollate

  17. from trackUtils.sort import Sort,track_draw_boxAndTrace,track_draw_trace_boxes,moving_average_wang,drawBoxTraceSimplied

  18. from trackUtils.sort_obb import OBB_Sort,obbTohbb,track_draw_all_boxes,track_draw_trace

  19. from obbUtils.shipUtils import OBB_infer,OBB_tracker,draw_obb,OBB_tracker_batch

  20. from utilsK.noParkingUtils import mixNoParking_road_postprocess

  21. from obbUtils.load_obb_model import load_model_decoder_OBB

  22. import numpy as np

  23. import torch,glob

  24. import tensorrt as trt

  25. from utilsK.masterUtils import get_needed_objectsIndex

  26. from copy import deepcopy

  27. from scipy import interpolate

  28. from utilsK.drownUtils import mixDrowing_water_postprocess

  29. #import warnings

  30. #warnings.filterwarnings("error")

  31. rainbows=[ [0,0,255],[0,255,0],[255,0,0],[255,0,255],[255,255,0],[255,129,0],[255,0,127],[127,255,0],[0,255,127],[0,127,255],[127,0,255],[255,127,255],[255,255,127],[127,255,255],[0,255,255],[255,127,255],[127,255,255],  [0,127,0],[0,0,127],[0,255,255]]

  32. def drawAllBox(preds,imgDraw,label_arraylist,rainbows,font):

  33.     for box in  preds:

  34.         #cls,conf,xyxy = box[0],box[5], box[1:5] 

  35.         #print('#'*20,'line47',box)

  36.         cls,conf,xyxy = box[5],box[4], box[0:4] ##2023.08.03,修改了格式

  37.         #print('#####line46 demo.py:', cls,conf,xyxy, len(label_arraylist),len(rainbows) )

  38.         imgDraw = draw_painting_joint(xyxy,imgDraw,label_arraylist[int(cls)],score=conf,color=rainbows[int(cls)%20],font=font,socre_location="leftBottom")

  39.     return imgDraw

  40. def get_drawPar(par):

  41.     labelnames = par['labelnames']

  42.     names=get_labelnames(labelnames)

  43.     mode_paras=par['detModelpara']

  44.     allowedList,allowedList_string=get_needed_objectsIndex(mode_paras) 

  45.     outfontsize=int(1080/1920*40);###

  46.     label_arraylist = get_label_arrays(names,rainbows,outfontsize=par['txtFontSize'],fontpath="../AIlib2/conf/platech.ttf")

  47.     drawPar={'names':names,'label_arraylist':label_arraylist,'rainbows':rainbows,'font': par['digitFont'],'allowedList':allowedList}

  48.     return drawPar

  49. def view_bar(num, total,time1,prefix='prefix'):

  50.     rate = num / total

  51.     time_n=time.time()

  52.     rate_num = int(rate * 30)

  53.     rate_nums = np.round(rate * 100)

  54.     r = '\r %s %d / %d [%s%s] %.2f s'%(prefix,num,total,  ">" * rate_num, " " * (30 - rate_num), time_n-time1 )

  55.     sys.stdout.write(r)

  56.     sys.stdout.flush()

  57.         

  58. 

  59. '''

  60.     多线程

  61. '''

  62. 

  63. def process_v1(frame):

  64.     #try:

  65.     print('demo.py beging to :',frame[8])

  66.     time00 = time.time()

  67.     H,W,C = frame[0][0].shape

  68.     

  69.     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])

  70.      

  71.     time11 = time.time()

  72.     image_array = p_result[1]

  73.     

  74.     cv2.imwrite(os.path.join('images/results/',frame[8] ) ,image_array)

  75.     bname = frame[8].split('.')[0]

  76.     if len(p_result)==5:

  77.         image_mask = p_result[4]

  78.         cv2.imwrite(os.path.join('images/results/',bname+'_mask.png' ) , (image_mask).astype(np.uint8))

  79.     

  80.     boxes=p_result[2]

  81.     with open(  os.path.join('images/results/',bname+'.txt' ),'w'  ) as fp:

  82.         for box in boxes:

  83.             box_str=[str(x) for x in box]

  84.             out_str=','.join(box_str)+'\n'

  85.             fp.write(out_str)

  86.     time22 = time.time()

  87.     print('%s,%d*%d,AI-process: %.1f,image save:%.1f , %s'%(frame[8],H,W, (time11 - time00) * 1000.0, (time22-time11)*1000.0,timeOut), boxes)

  88.     return 'success'

  89.     #except Exception as e:

  90.     #    return 'failed:'+str(e)        

  91. def process_video(video,par0,mode='detSeg'):

  92.     cap=cv2.VideoCapture(video)

  93.     if not  cap.isOpened():

  94.         print('#####error url:',video)

  95.         return False  

  96.     bname=os.path.basename(video).split('.')[0]        

  97.     fps = int(cap.get(cv2.CAP_PROP_FPS)+0.5)

  98.     width = int(cap.get(cv2.CAP_PROP_FRAME_WIDTH )+0.5)

  99.     height = int(cap.get(cv2.CAP_PROP_FRAME_HEIGHT)+0.5)

  100.     framecnt=int(cap.get(7)+0.5)

  101.     save_path_AI = os.path.join(par0['outpth'],os.path.basename(video))

  102.     problem_image_dir= os.path.join(  par0['outpth'], 'probleImages'   )

  103.     os.makedirs(problem_image_dir,exist_ok=True)

  104.     vid_writer_AI = cv2.VideoWriter(save_path_AI, cv2.VideoWriter_fourcc(*'mp4v'), fps, (width,height))

  105.     num=0

  106.     iframe=0;post_results=[];fpsample=30*10

  107.     

  108.     imgarray_list = []; iframe_list = []

  109.     

  110.     patch_cnt = par0['trackPar']['patchCnt']

  111.     ##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。

  112.     trackPar={'det_cnt':10,'windowsize':29 }

  113. 

  114. 

  115.     ##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.

  116.     while cap.isOpened():

  117.         ret, imgarray = cap.read() #读取摄像头画面

  118.         iframe +=1

  119.         if not ret:break

  120.         if mode=='detSeg':

  121.             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'])

  122.         elif mode == 'track':

  123.             #sampleCount=10

  124.             imgarray_list.append( imgarray )

  125.             iframe_list.append(iframe )    

  126.             if iframe%patch_cnt==0:

  127.                 time_patch0 = time.time()

  128.                 #retResults,timeInfos = AI_det_track_batch(imgarray_list, iframe_list ,par0['modelPar'],par0['processPar'],par0['sort_tracker'] ,par0['trackPar'],segPar=par0['segPar'])

  129.                 retResults,timeInfos = AI_det_track_batch_N(imgarray_list, iframe_list ,par0['modelList'],par0['postProcess'],par0['sort_tracker'],par0['trackPar']) 

  130.                 print('###line111:',retResults[2])

  131.                 ###需要保存成一个二维list，每一个list是一帧检测结果。

  132.                 ###track_det_result 内容格式：x1, y1, x2, y2, conf, cls,iframe,trackId

  133.                 time_patch2 = time.time()                    

  134.                 frame_min = iframe_list[0];frame_max=iframe_list[-1]

  135.                 for iiframe in range(frame_min,frame_max+1):

  136.                     img_draw = imgarray_list[ iiframe- frame_min  ]

  137.                     img_draw = drawBoxTraceSimplied(retResults[1] ,iiframe, img_draw,rainbows=par0['drawPar']['rainbows'],boxFlag=True,traceFlag=True,names=par0['drawPar']['names'] )

  138.                     ret = vid_writer_AI.write(img_draw)        

  139.                     view_bar(iiframe, framecnt,time.time(),prefix=os.path.basename(video))

  140.                 imgarray_list=[];iframe_list=[]     

  141.         elif mode =='obbTrack':

  142.             imgarray_list.append( imgarray )

  143.             iframe_list.append(iframe )    

  144.             if iframe%patch_cnt==0:

  145.                 time_patch0 = time.time()

  146.                     

  147.                 track_det_results,  timeInfos = OBB_tracker_batch(imgarray_list,iframe_list,par0['modelPar'],par0['obbModelPar'],par0['sort_tracker'],par0['trackPar'],segPar=None)

  148.                 print(  timeInfos )

  149.                 

  150.                 #对结果画图

  151.                 track_det_np = track_det_results[1]

  152.                 frame_min = iframe_list[0];frame_max=iframe_list[-1]

  153.                 for iiframe in range(frame_min,frame_max+1):

  154.                     img_draw = imgarray_list[ iiframe- frame_min  ]

  155.                     

  156.                     if len( track_det_results[2][  iiframe- frame_min]) > 0:

  157.                         img_draw =  draw_obb( track_det_results[2][iiframe- frame_min ] ,img_draw,par0['drawPar'])

  158.                         if  True: 

  159.                             frameIdex=12;trackIdex=13;

  160.                             boxes_oneFrame = track_det_np[ track_det_np[:,frameIdex]==iiframe   ]

  161.                             

  162.                             ###在某一帧上，画上轨迹

  163.                             track_ids = boxes_oneFrame[:,trackIdex].tolist()

  164.                             boxes_before_oneFrame = track_det_np[ track_det_np[:,frameIdex]<=iiframe   ]

  165.                             for trackId in track_ids:

  166.                                 boxes_before_oneFrame_oneId = boxes_before_oneFrame[boxes_before_oneFrame[:,trackIdex]==trackId]

  167.                                 xcs = boxes_before_oneFrame_oneId[:,8]

  168.                                 ycs = boxes_before_oneFrame_oneId[:,9]

  169.                                 [cv2.line(img_draw, (  int(xcs[i]) , int(ycs[i]) ), 

  170.                                     ( int(xcs[i+1]),int(ycs[i+1])  ),(255,0,0), thickness=2) 

  171.                                     for i,_ in  enumerate(xcs)  if i < len(xcs)-1 ]

  172. 

  173.                     ret = vid_writer_AI.write(img_draw)   

  174.                    

  175.                 #sys.exit(0)

  176.                 #print('vide  writer ret:',ret)

  177.                 imgarray_list=[];iframe_list=[]     

  178.   

  179.             view_bar(iframe, framecnt,time.time(),prefix=os.path.basename(video))    

  180.             

  181.         else:

  182.             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'])

  183.             

  184. 

  185.         if  mode not in [ 'track','obbTrack']:   

  186.             image_array = p_result[1];num+=1

  187.             ret = vid_writer_AI.write(image_array)    

  188.             view_bar(num, framecnt,time.time(),prefix=os.path.basename(video))

  189.             ##每隔 fpsample帧处理一次，如果有问题就保存图片  

  190.             if (iframe % fpsample == 0) and (len(post_results)>0) :

  191.                 parImage=save_problem_images(post_results,iframe,par0['names'],streamName=bname,outImaDir=problem_image_dir,imageTxtFile=False)

  192.                 post_results=[]   

  193.     

  194.             if len(p_result[2] )>0:                

  195.                 post_results.append(p_result)

  196.  

  197.     vid_writer_AI.release();       

  198.    

  199. def det_track_demo_N(business ):

  200.     ####河道巡检的跟踪模型参数

  201.     if opt['business'] == 'river' or opt['business'] == 'river2' or opt['business'] == 'riverT': 

  202.         from utilsK.queRiver import riverDetSegMixProcess_N

  203.         par={

  204.             'device':'0', ###显卡号,如果用TRT模型，只支持0（单显卡）

  205.             'labelnames':"../AIlib2/weights/conf/%s/labelnames.json"%(opt['business']), ###检测类别对照表

  206.             'max_workers':1, ###并行线程数

  207.             'trackPar':{'sort_max_age':2,'sort_min_hits':3,'sort_iou_thresh':0.2,'det_cnt':10,'windowsize':29,'patchCnt':100},

  208.             'postProcess':{'function':riverDetSegMixProcess_N,'pars':{'slopeIndex':[1,3,4,7], 'riverIou':0.1}}, #分割和检测混合处理的函数         

  209.             'models':

  210.                 [

  211.                     {

  212.                         'weight':"../weights/%s/AIlib2/%s/yolov5_%s_fp16.engine"%(opt['gpu'], opt['business'] ,opt['gpu']  ),###检测模型路径

  213.                         #'weight':'../AIlib2/weights/conf/%s/yolov5.pt'%(opt['business']  ),

  214.                         'name':'yolov5',

  215.                         'model':yolov5Model,

  216.                         'par':{ 'half':True,'device':'cuda:0' ,'conf_thres':0.25,'iou_thres':0.45,'allowedList':[0,1,2,3],'segRegionCnt':1, 'trtFlag_det':False,'trtFlag_seg':False, "score_byClass":{"0":0.25,"1":0.3,"2":0.3,"3":0.3  } },       

  217.                     },

  218.                     {

  219.                         'weight':"../weights/%s/AIlib2/%s/stdc_360X640_%s_fp16.engine"%(opt['gpu'], opt['business'] ,opt['gpu']  ),###检测模型路径

  220.                         #'weight':'../AIlib2/weights/conf/%s/stdc_360X640.pth'%(opt['business']  ),

  221.                         'par':{

  222.                             'modelSize':(640,360),'mean':(0.485, 0.456, 0.406),'std' :(0.229, 0.224, 0.225),'numpy':False, 'RGB_convert_first':True,'seg_nclass':2},###分割模型预处理参数

  223.                         'model':stdcModel,

  224.                         'name':'stdc'        

  225.                     }

  226.                         

  227.                 ],

  228.             '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

  229.             'postFile': '../AIlib2/weights/conf/%s/para.json'%( opt['business'] ),###后处理参数文件

  230.             'txtFontSize':80,###文本字符的大小

  231.             'digitFont': { 'line_thickness':2,'boxLine_thickness':1, 'fontSize':1.0,'waterLineColor':(0,255,255),'waterLineWidth':3},###显示框、线设置

  232.             #'testImgPath':'images/videos/river',###测试图像的位置

  233.             'testImgPath':'images/%s'%(opt['business']),###测试图像的位置

  234.             'testOutPath':'images/results/',###输出测试图像位置

  235.             

  236.         }

  237. 

  238.     if opt['business'] == 'highWay2':

  239.         from  segutils.trafficUtils  import tracfficAccidentMixFunction_N 

  240.         par={

  241.             'device':'0', ###显卡号,如果用TRT模型，只支持0（单显卡）

  242.             'labelnames':"../AIlib2/weights/conf/%s/labelnames.json"%(  opt['business']  ), ###检测类别对照表

  243.             'max_workers':1, ###并行线程数

  244.             'trackPar':{'sort_max_age':2,'sort_min_hits':3,'sort_iou_thresh':0.2,'det_cnt':5,'windowsize':29,'patchCnt':100},

  245.             'postProcess':{'function':tracfficAccidentMixFunction_N,

  246.                 'pars':{ 'RoadArea': 16000, 'vehicleArea': 10, 'roadVehicleAngle': 15, 'speedRoadVehicleAngleMax': 75,'radius': 50 , 'roundness': 1.0, 'cls': 10, 'vehicleFactor': 0.1,'cls':9, 'confThres':0.25,'roadIou':0.6,'vehicleFlag':False,'distanceFlag': False }

  247.                 },

  248.             'models':

  249.                 [

  250.                     {

  251.                         'weight':"../weights/%s/AIlib2/%s/yolov5_%s_fp16.engine"%(opt['gpu'], opt['business'] ,opt['gpu']  ),###检测模型路径

  252.                         #'weight':'../AIlib2/weights/conf/%s/yolov5.pt'%(opt['business']  ),

  253.                         'name':'yolov5',

  254.                         'model':yolov5Model,

  255.                         'par':{ 'half':True,'device':'cuda:0' ,'conf_thres':0.25,'iou_thres':0.45,'allowedList':[0,1,2,3],'segRegionCnt':1, 'trtFlag_det':False,'trtFlag_seg':False, "score_byClass":{"0":0.25,"1":0.3,"2":0.3,"3":0.3  } },       

  256.                     },

  257.                     {

  258.                         #'weight':"../weights/%s/AIlib2/%s/dmpr_%s.engine"%(opt['gpu'], opt['business'] ,opt['gpu']  ),###检测模型路径

  259.                         'weight':'../AIlib2/weights/conf/%s/stdc_360X640.pth'%(opt['business']  ),

  260.                         'par':{

  261.                             '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,'seg_nclass':3},###分割模型预处理参数

  262.                         'model':stdcModel,

  263.                         'name':'stdc'        

  264.                     }

  265.                         

  266.                 ],

  267.                 

  268. 

  269.             '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

  270.             'postFile': '../AIlib2/weights/conf/%s/para.json'%(opt['business'] ),###后处理参数文件

  271.             'txtFontSize':20,###文本字符的大小

  272.             'digitFont': { 'line_thickness':2,'boxLine_thickness':1, 'fontSize':0.5,'waterLineColor':(0,255,255),'segLineShow':True,'waterLineWidth':2},###显示框、线设置

  273.             #'testImgPath':'images/trafficAccident/8.png',###测试图像的位置

  274.             #'testImgPath':'/home/chenyukun/777-7-42.mp4',###测试图像的位置

  275.             'testImgPath':'/mnt/thsw2/DSP2/weights/highWay2/images/',

  276.             'testOutPath':'images/results/',###输出测试图像位置

  277.         }

  278.         par['postProcess']['pars']['modelSize'] = par['models'][1]['par']['modelSize']

  279.           

  280.     if opt['business'] == 'noParking':

  281.         from utilsK.noParkingUtils import mixNoParking_road_postprocess_N

  282.         par={

  283.             'device':'0', ###显卡号,如果用TRT模型，只支持0（单显卡）

  284.             'labelnames':"../AIlib2/weights/conf/%s/labelnames.json"%(  opt['business']  ), ###检测类别对照表

  285.             'max_workers':1, ###并行线程数

  286.             'trackPar':{'sort_max_age':2,'sort_min_hits':3,'sort_iou_thresh':0.2,'det_cnt':10,'windowsize':29,'patchCnt':100},

  287.             'postProcess':{'function':mixNoParking_road_postprocess_N,

  288.                         'pars': { 'roundness': 0.3, 'cls': 9,  'laneArea': 10, 'laneAngleCha': 5 ,'RoadArea': 16000,'fitOrder':2}

  289.                         } ,

  290.             'models':

  291.                 [

  292.                     {

  293.                         'weight':"../weights/%s/AIlib2/%s/yolov5_%s_fp16.engine"%(opt['gpu'], opt['business'] ,opt['gpu']  ),###检测模型路径

  294.                         #'weight':'../AIlib2/weights/conf/%s/yolov5.pt'%(opt['business']  ),

  295.                         'name':'yolov5',

  296.                         'model':yolov5Model,

  297.                         'par':{ 'half':True,'device':'cuda:0' ,'conf_thres':0.25,'iou_thres':0.45,'allowedList':[0,1,2,3],'segRegionCnt':1, 'trtFlag_det':False,'trtFlag_seg':False, "score_byClass":{"0":0.25,"1":0.3,"2":0.3,"3":0.3  } },       

  298.                     },

  299.                     {

  300.                         #'weight':"../weights/%s/AIlib2/%s/dmpr_%s.engine"%(opt['gpu'], opt['business'] ,opt['gpu']  ),###检测模型路径

  301.                         'weight':'../AIlib2/weights/conf/%s/stdc_360X640.pth'%(opt['business']  ),

  302.                         'par':{

  303.                             '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,'seg_nclass':4},###分割模型预处理参数

  304.                         'model':stdcModel,

  305.                         'name':'stdc'        

  306.                     }

  307.                         

  308.                 ],

  309.             '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

  310.             'postFile': '../AIlib2/weights/conf/%s/para.json'%('highWay2' ),###后处理参数文件

  311.             'txtFontSize':20,###文本字符的大小

  312.             'digitFont': { 'line_thickness':2,'boxLine_thickness':1, 'fontSize':1.0,'waterLineColor':(0,255,255),'segLineShow':True,'waterLineWidth':2},###显示框、线设置

  313.             'testImgPath':'images/noParking/',###测试图像的位置

  314.             'testOutPath':'images/results/',###输出测试图像位置

  315.         }

  316.         par['postProcess']['pars']['modelSize'] = par['models'][1]['par']['modelSize']

  317.         

  318.     if opt['business'] == 'drowning':

  319.         from utilsK.drownUtils import mixDrowing_water_postprocess_N

  320.         par={

  321.             'device':'0', ###显卡号,如果用TRT模型，只支持0（单显卡）

  322.             'labelnames':"../AIlib2/weights/conf/%s/labelnames.json"%(  opt['business']  ), ###检测类别对照表

  323.             'max_workers':1, ###并行线程数

  324.             'trackPar':{'sort_max_age':2,'sort_min_hits':3,'sort_iou_thresh':0.2,'det_cnt':10,'windowsize':29,'patchCnt':100},

  325.             'postProcess':{'function':mixDrowing_water_postprocess_N,

  326.                         'pars':{ }},

  327.             'models':

  328.                 [

  329.                     {

  330.                         'weight':"../weights/%s/AIlib2/%s/yolov5_%s_fp16.engine"%(opt['gpu'], opt['business'] ,opt['gpu']  ),###检测模型路径

  331.                         #'weight':'../AIlib2/weights/conf/%s/yolov5.pt'%(opt['business']  ),

  332.                         'name':'yolov5',

  333.                         'model':yolov5Model,

  334.                         'par':{ 'half':True,'device':'cuda:0' ,'conf_thres':0.25,'iou_thres':0.45,'allowedList':[0,1,2,3],'segRegionCnt':1, 'trtFlag_det':False,'trtFlag_seg':False, "score_byClass":{"0":0.25,"1":0.3,"2":0.3,"3":0.3  } },       

  335.                     },

  336.                     {

  337.                         #'weight':"../weights/%s/AIlib2/%s/dmpr_%s.engine"%(opt['gpu'], opt['business'] ,opt['gpu']  ),###检测模型路径

  338.                         'weight':'../AIlib2/weights/conf/%s/stdc_360X640.pth'%(opt['business']  ),

  339.                         'par':{

  340.                             '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,'seg_nclass':2},###分割模型预处理参数

  341.                         'model':stdcModel,

  342.                         'name':'stdc'        

  343.                     }

  344.                         

  345.                 ],

  346.             '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    

  347.             'postFile': '../AIlib2/weights/conf/%s/para.json'%('highWay2' ),###后处理参数文件

  348.             'txtFontSize':20,###文本字符的大小

  349.             'digitFont': { 'line_thickness':2,'boxLine_thickness':1, 'fontSize':1.0,'waterLineColor':(0,255,255),'segLineShow':True,'waterLineWidth':2},###显示框、线设置

  350.             'testImgPath':'images/drowning/',###测试图像的位置

  351.             'testOutPath':'images/results/',###输出测试图像位置

  352.         }

  353.         par['postProcess']['pars']['modelSize'] = par['models'][1]['par']['modelSize']

  354.         

  355.     if opt['business'] == 'cityMangement2':

  356.         from DMPR import DMPRModel

  357.         from DMPRUtils.jointUtil import  dmpr_yolo_stdc

  358. 

  359.         par={

  360.             'device':'0', ###显卡号,如果用TRT模型，只支持0（单显卡）

  361.             'labelnames':"../AIlib2/weights/conf/%s/labelnames.json"%(opt['business']), ###检测类别对照表

  362.             'max_workers':1, ###并行线程数

  363.             'trackPar':{'sort_max_age':2,'sort_min_hits':3,'sort_iou_thresh':0.2,'det_cnt':10,'windowsize':29,'patchCnt':100},     

  364.             'postProcess':{

  365.                     'function':dmpr_yolo_stdc,

  366.                     'pars':{'carCls':0 ,'illCls':3,'scaleRatio':0.5,'border':80}

  367.                     }, 

  368.             'models':

  369.                 [

  370.                     {

  371.                         #'weight':"../weights/%s/AIlib2/%s/yolov5_%s_fp16.engine"%(opt['gpu'], opt['business'] ,opt['gpu']  ),###检测模型路径

  372.                         'weight':'../AIlib2/weights/conf/%s/yolov5.pt'%(opt['business']  ),

  373.                         'name':'yolov5',

  374.                         'model':yolov5Model,

  375.                         'par':{ 'half':True,'device':'cuda:0' ,'conf_thres':0.25,'iou_thres':0.45,'allowedList':[0,1,2,3],'segRegionCnt':1, 'trtFlag_det':False,'trtFlag_seg':False, "score_byClass":{"0":0.25,"1":0.3,"2":0.3,"3":0.3  } }

  376.                     

  377.                     },

  378.                     {

  379.                         #'weight':"../weights/%s/AIlib2/%s/dmpr_%s.engine"%(opt['gpu'], opt['business'] ,opt['gpu']  ),###检测模型路径

  380.                         'weight':'../AIlib2/weights/conf/%s/dmpr.pth'%(opt['business']  ),

  381.                         'par':{

  382.                             'depth_factor':32,'NUM_FEATURE_MAP_CHANNEL':6,'dmpr_thresh':0.3, 'dmprimg_size':640,

  383.                             'name':'dmpr'

  384.                              },

  385.                         'model':DMPRModel,     

  386.                         'name':'dmpr'

  387.                     }

  388.             ]   ,

  389.             '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

  390.             'postFile': '../AIlib2/weights/conf/%s/para.json'%(opt['business'] ),###后处理参数文件

  391.             'txtFontSize':20,###文本字符的大小

  392.             'digitFont': { 'line_thickness':2,'boxLine_thickness':1, 'fontSize':1.0,'waterLineColor':(0,255,255),'segLineShow':True,'waterLineWidth':2},###显示框、线设置

  393.             #'testImgPath':'/mnt/thsw2/DSP2/demoImages/illParking',###测试图像的位置

  394.             'testImgPath':'/mnt/thsw2/DSP2/weights/cityMangement2_0916/images/input',

  395.             #'testImgPath':'images/cityMangement/',

  396.             'testOutPath':'images/results/',###输出测试图像位置

  397.         }

  398.         

  399.     if opt['business'] == 'cityMangement3':

  400.         from DMPR import DMPRModel

  401.         from DMPRUtils.jointUtil import  dmpr_yolo_stdc

  402.         par={

  403.                 'device':'0', ###显卡号,如果用TRT模型，只支持0（单显卡）

  404.                 'labelnames':"../AIlib2/weights/conf/%s/labelnames.json"%(opt['business']), ###检测类别对照表

  405.                 'max_workers':1, ###并行线程数

  406.                 'trackPar':{'sort_max_age':2,'sort_min_hits':3,'sort_iou_thresh':0.2,'det_cnt':10,'windowsize':29,'patchCnt':100},       

  407.                 'postProcess':{

  408.                     'function':dmpr_yolo_stdc,

  409.                     'pars':{'carCls':0 ,'illCls':3,'scaleRatio':0.5,'border':80}

  410.                     }, 

  411.                 'models':[

  412.                         {

  413.                             'weight':"../weights/%s/AIlib2/%s/yolov5_%s_fp16.engine"%(opt['gpu'], opt['business'] ,opt['gpu']  ),###检测模型路径

  414.                             #'weight':'../AIlib2/weights/conf/%s/yolov5.pt'%(opt['business']  ),

  415.                             'name':'yolov5',

  416.                             'model':yolov5Model,

  417.                             'par':{ 'half':True,'device':'cuda:0' ,'conf_thres':0.25,'iou_thres':0.45,'allowedList':[0,1,2,3],'segRegionCnt':1, 'trtFlag_det':False,'trtFlag_seg':False, "score_byClass":{"0":0.25,"1":0.3,"2":0.3,"3":0.3  } }

  418.                         

  419.                         },

  420.                         {

  421.                             'weight':"../weights/%s/AIlib2/%s/dmpr_%s.engine"%(opt['gpu'], opt['business'] ,opt['gpu']  ),###DMPR模型路径

  422.                             #'weight':'../AIlib2/weights/conf/%s/dmpr.pth'%(opt['business']  ),

  423.                             'par':{

  424.                                 'depth_factor':32,'NUM_FEATURE_MAP_CHANNEL':6,'dmpr_thresh':0.1, 'dmprimg_size':640,

  425.                                 'name':'dmpr'

  426.                                  },

  427.                             'model':DMPRModel,     

  428.                             'name':'dmpr'

  429.                         },

  430.                         {

  431.                             'weight':"../weights/%s/AIlib2/%s/stdc_360X640_%s_fp16.engine"%(opt['gpu'], opt['business'] ,opt['gpu']  ),###分割模型路径

  432.                           

  433.                             #'weight':'../AIlib2/weights/conf/%s/stdc_360X640.pth'%(opt['business']  ),

  434.                             'par':{

  435.                                 '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,'seg_nclass':2},###分割模型预处理参数

  436.                             'model':stdcModel,

  437.                             'name':'stdc'        

  438.                             

  439.                         } 

  440.                     ],

  441.                 

  442.                 '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

  443.                 'postFile': '../AIlib2/weights/conf/%s/para.json'%(opt['business'] ),###后处理参数文件

  444.                 'txtFontSize':20,###文本字符的大小

  445.                 'digitFont': { 'line_thickness':2,'boxLine_thickness':1, 'fontSize':1.0,'waterLineColor':(0,255,255),'segLineShow':True,'waterLineWidth':2},###显示框、线设置

  446.                 'testImgPath':'images/%s/'%(opt['business']),

  447.                 #'testImgPath':'/mnt/thsw2/DSP2/weights/cityMangement3/images/',

  448.                 'testOutPath':'images/results/',###输出测试图像位置    

  449.             }

  450.     

  451.     if opt['business'] == 'forest2':

  452.         

  453.         par={

  454.             'device':'0', ###显卡号,如果用TRT模型，只支持0（单显卡）

  455.             'labelnames':"../AIlib2/weights/conf/forest2/labelnames.json", ###检测类别对照表

  456.             'max_workers':1, ###并行线程数

  457.             'trackPar':{'sort_max_age':2,'sort_min_hits':3,'sort_iou_thresh':0.2,'det_cnt':10,'windowsize':29,'patchCnt':100},

  458.             'postProcess':{'function':default_mix,'pars':{ }},

  459.             'models':

  460.                 [

  461.                     {

  462.                         'weight':"../weights/%s/AIlib2/%s/yolov5_%s_fp16.engine"%(opt['gpu'], opt['business'] ,opt['gpu']  ),###检测模型路径

  463.                         #'weight':'../AIlib2/weights/conf/%s/yolov5.pt'%(opt['business']  ),

  464.                         'name':'yolov5',

  465.                         'model':yolov5Model,

  466.                         'par':{ 'half':True,'device':'cuda:0' ,'conf_thres':0.25,'iou_thres':0.45,'allowedList':[0,1,2,3],'segRegionCnt':1, 'trtFlag_det':False,'trtFlag_seg':False, "score_byClass":{"0":0.25,"1":0.3,"2":0.3,"3":0.3  } },       

  467.                     }

  468.                     

  469.                         

  470.                 ],

  471.             'detModelpara':[{"id":str(x),"config":{"k1":"v1","k2":"v2"}} for x in []  ],

  472.             'postFile': '../AIlib2/weights/conf/forest/para.json',###后处理参数文件

  473.             'txtFontSize':80,###文本字符的大小

  474.             'digitFont': { 'line_thickness':2,'boxLine_thickness':1, 'fontSize':1.0,'waterLineColor':(0,255,255),'waterLineWidth':3},###显示框、线设置

  475.             'testImgPath':'../AIdemo2/images/forest2/',###测试图像的位置

  476.             'testOutPath':'images/results/',###输出测试图像位置

  477.         }

  478.      ###车辆巡检参数

  479.     if opt['business'] == 'vehicle': 

  480.         par={

  481.             'device':'0', ###显卡号,如果用TRT模型，只支持0（单显卡）

  482.             'labelnames':"../AIlib2/weights/conf/vehicle/labelnames.json", ###检测类别对照表

  483.             'max_workers':1, ###并行线程数

  484.             'trackPar':{'sort_max_age':2,'sort_min_hits':3,'sort_iou_thresh':0.2,'det_cnt':10,'windowsize':29,'patchCnt':100},

  485.             'postProcess':{'function':default_mix,'pars':{ }},

  486.             'models':

  487.                 [

  488.                     {

  489.                         'weight':"../weights/%s/AIlib2/%s/yolov5_%s_fp16.engine"%(opt['gpu'], opt['business'] ,opt['gpu']  ),###检测模型路径

  490.                         #'weight':'../AIlib2/weights/conf/%s/yolov5.pt'%(opt['business']  ),

  491.                         'name':'yolov5',

  492.                         'model':yolov5Model,

  493.                         'par':{ 'half':True,'device':'cuda:0' ,'conf_thres':0.25,'iou_thres':0.45,'allowedList':[0,1,2,3],'segRegionCnt':1, 'trtFlag_det':False,'trtFlag_seg':False, "score_byClass":{"0":0.25,"1":0.3,"2":0.3,"3":0.3  } },       

  494.                     }

  495.                     

  496.                         

  497.                 ], 

  498.             '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

  499.             'postFile': '../AIlib2/weights/conf/vehicle/para.json',###后处理参数文件

  500.             'txtFontSize':40,###文本字符的大小

  501.             'digitFont': { 'line_thickness':2,'boxLine_thickness':1, 'fontSize':1.0,'waterLineColor':(0,255,255),'waterLineWidth':3},###显示框、线设置

  502.             'testImgPath':'images/vehicle/',###测试图像的位置

  503.             'testOutPath':'images/results/',###输出测试图像位置

  504.         }    

  505.      ###行人检测模型

  506.     if opt['business'] == 'pedestrian': 

  507.         par={

  508.             'device':'0', ###显卡号,如果用TRT模型，只支持0（单显卡）

  509.             'labelnames':"../AIlib2/weights/conf/pedestrian/labelnames.json", ###检测类别对照表

  510.             'max_workers':1, ###并行线程数

  511.             'trackPar':{'sort_max_age':2,'sort_min_hits':3,'sort_iou_thresh':0.2,'det_cnt':10,'windowsize':29,'patchCnt':100},

  512.             'postProcess':{'function':default_mix,'pars':{ }},

  513.             'models':

  514.                 [

  515.                     {

  516.                         'weight':"../weights/%s/AIlib2/%s/yolov5_%s_fp16.engine"%(opt['gpu'], opt['business'] ,opt['gpu']  ),###检测模型路径

  517.                         #'weight':'../AIlib2/weights/conf/%s/yolov5.pt'%(opt['business']  ),

  518.                         'name':'yolov5',

  519.                         'model':yolov5Model,

  520.                         'par':{ 'half':True,'device':'cuda:0' ,'conf_thres':0.25,'iou_thres':0.45,'allowedList':[0,1,2,3],'segRegionCnt':1, 'trtFlag_det':False,'trtFlag_seg':False, "score_byClass":{"0":0.25,"1":0.3,"2":0.3,"3":0.3  } },       

  521.                     }

  522.                     

  523.                         

  524.                 ], 

  525.           

  526.             '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

  527.             'postFile': '../AIlib2/weights/conf/pedestrian/para.json',###后处理参数文件

  528.             'txtFontSize':40,###文本字符的大小

  529.             'digitFont': { 'line_thickness':2,'boxLine_thickness':1, 'fontSize':1.0,'waterLineColor':(0,255,255),'waterLineWidth':3},###显示框、线设置

  530.             'testImgPath':'../AIdemo2/images/pedestrian/',###测试图像的位置

  531.             'testOutPath':'images/results/',###输出测试图像位置

  532.         }

  533.     if opt['business'] == 'smogfire': 

  534.         par={

  535.             'device':'0', ###显卡号,如果用TRT模型，只支持0（单显卡）

  536.             'labelnames':"../AIlib2/weights/conf/smogfire/labelnames.json", ###检测类别对照表

  537.             'max_workers':1, ###并行线程数

  538.             'trackPar':{'sort_max_age':2,'sort_min_hits':3,'sort_iou_thresh':0.2,'det_cnt':10,'windowsize':29,'patchCnt':100},

  539.             'postProcess':{'function':default_mix,'pars':{ }},

  540.             'models':

  541.                 [

  542.                     {

  543.                         'weight':"../weights/%s/AIlib2/%s/yolov5_%s_fp16.engine"%(opt['gpu'], opt['business'] ,opt['gpu']  ),###检测模型路径

  544.                         #'weight':'../AIlib2/weights/conf/%s/yolov5.pt'%(opt['business']  ),

  545.                         'name':'yolov5',

  546.                         'model':yolov5Model,

  547.                         'par':{ 'half':True,'device':'cuda:0' ,'conf_thres':0.25,'iou_thres':0.45,'allowedList':[0,1,2,3],'segRegionCnt':1, 'trtFlag_det':False,'trtFlag_seg':False, "score_byClass":{"0":0.25,"1":0.3,"2":0.3,"3":0.3  } },       

  548.                     }

  549.                 ],  

  550.             '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

  551.             'postFile': '../AIlib2/weights/conf/smogfire/para.json',###后处理参数文件

  552.             'txtFontSize':40,###文本字符的大小

  553.             'digitFont': { 'line_thickness':2,'boxLine_thickness':1, 'fontSize':1.0,'waterLineColor':(0,255,255),'waterLineWidth':3},###显示框、线设置

  554.             'testImgPath':'../AIdemo2/images/smogfire/',###测试图像的位置

  555.             'testOutPath':'images/results/',###输出测试图像位置

  556.         }

  557.     ###钓鱼游泳检测  

  558.     if opt['business'] == 'AnglerSwimmer': 

  559.         par={

  560.             'device':'0', ###显卡号,如果用TRT模型，只支持0（单显卡）

  561.             'labelnames':"../AIlib2/weights/conf/AnglerSwimmer/labelnames.json", ###检测类别对照表

  562.             'max_workers':1, ###并行线程数

  563.             'trackPar':{'sort_max_age':2,'sort_min_hits':3,'sort_iou_thresh':0.2,'det_cnt':10,'windowsize':29,'patchCnt':100},

  564.             'postProcess':{'function':default_mix,'pars':{ }},

  565.             'models':

  566.                 [

  567.                     {

  568.                         'weight':"../weights/%s/AIlib2/%s/yolov5_%s_fp16.engine"%(opt['gpu'], opt['business'] ,opt['gpu']  ),###检测模型路径

  569.                         #'weight':'../AIlib2/weights/conf/%s/yolov5.pt'%(opt['business']  ),

  570.                         'name':'yolov5',

  571.                         'model':yolov5Model,

  572.                         'par':{ 'half':True,'device':'cuda:0' ,'conf_thres':0.25,'iou_thres':0.45,'allowedList':[0,1,2,3],'segRegionCnt':1, 'trtFlag_det':False,'trtFlag_seg':False, "score_byClass":{"0":0.25,"1":0.3,"2":0.3,"3":0.3  } },       

  573.                     }

  574.                 ],  

  575.             '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

  576.             'postFile': '../AIlib2/weights/conf/AnglerSwimmer/para.json',###后处理参数文件

  577.             'txtFontSize':40,###文本字符的大小

  578.             'digitFont': { 'line_thickness':2,'boxLine_thickness':1, 'fontSize':1.0,'waterLineColor':(0,255,255),'waterLineWidth':3},###显示框、线设置

  579.             'testImgPath':'../AIdemo2/images/AnglerSwimmer/',###测试图像的位置

  580.             'testOutPath':'images/results/',###输出测试图像位置

  581.         }

  582.     #单独的“坑槽”检测    

  583.     if opt['business'] == 'pothole':

  584.         

  585.         par={

  586.             'device':'0', ###显卡号,如果用TRT模型，只支持0（单显卡）

  587.             'labelnames':"../AIlib2/weights/conf/pothole/labelnames.json", ###检测类别对照表

  588.             'max_workers':1, ###并行线程数

  589.             'trackPar':{'sort_max_age':2,'sort_min_hits':3,'sort_iou_thresh':0.2,'det_cnt':10,'windowsize':29,'patchCnt':100},

  590.             'postProcess':{'function':default_mix,'pars':{ }},

  591.             'models':

  592.                 [

  593.                     {

  594.                         'weight':"../weights/%s/AIlib2/%s/yolov5_%s_fp16.engine"%(opt['gpu'], opt['business'] ,opt['gpu']  ),###检测模型路径

  595.                         #'weight':'../AIlib2/weights/conf/%s/yolov5.pt'%(opt['business']  ),

  596.                         'name':'yolov5',

  597.                         'model':yolov5Model,

  598.                         'par':{ 'half':True,'device':'cuda:0' ,'conf_thres':0.25,'iou_thres':0.45,'allowedList':[0,1,2,3],'segRegionCnt':1, 'trtFlag_det':False,'trtFlag_seg':False, "score_byClass":{"0":0.25,"1":0.3,"2":0.3,"3":0.3  } },       

  599.                     }

  600.                     

  601.                         

  602.                 ],

  603.             'detModelpara':[{"id":str(x),"config":{"k1":"v1","k2":"v2"}} for x in []  ],

  604.             'postFile': '../AIlib2/weights/conf/forest/para.json',###后处理参数文件

  605.             'txtFontSize':80,###文本字符的大小

  606.             'digitFont': { 'line_thickness':2,'boxLine_thickness':1, 'fontSize':1.0,'waterLineColor':(0,255,255),'waterLineWidth':3},###显示框、线设置

  607.             'testImgPath':'../AIdemo2/images/pothole/',###测试图像的位置

  608.             'testOutPath':'images/results/',###输出测试图像位置

  609.         }

  610.         

  611.     ###航道应急，做落水人员检测, channelEmergency    

  612.     if opt['business'] == 'channelEmergency': 

  613.         par={

  614.             'device':'0', ###显卡号,如果用TRT模型，只支持0（单显卡）

  615.             'labelnames':"../AIlib2/weights/conf/channelEmergency/labelnames.json", ###检测类别对照表

  616.             'max_workers':1, ###并行线程数

  617.             'trackPar':{'sort_max_age':2,'sort_min_hits':3,'sort_iou_thresh':0.2,'det_cnt':10,'windowsize':29,'patchCnt':100},

  618.             'postProcess':{'function':default_mix,'pars':{ }},

  619.             'models':

  620.                 [

  621.                     {

  622.                         'weight':"../weights/%s/AIlib2/%s/yolov5_%s_fp16.engine"%(opt['gpu'], opt['business'] ,opt['gpu']  ),###检测模型路径

  623.                         #'weight':'../AIlib2/weights/conf/%s/yolov5.pt'%(opt['business']  ),

  624.                         'name':'yolov5',

  625.                         'model':yolov5Model,

  626.                         'par':{ 'half':True,'device':'cuda:0' ,'conf_thres':0.25,'iou_thres':0.45,'allowedList':[0,1,2,3],'segRegionCnt':1, 'trtFlag_det':False,'trtFlag_seg':False, "score_byClass":{"0":0.25,"1":0.3,"2":0.3,"3":0.3  } },       

  627.                     }

  628.                 ],  

  629.             'detModelpara':[{"id":str(x),"config":{"k1":"v1","k2":"v2"}} for x in []  ],###控制哪些检测类别显示、输出

  630.             

  631.            

  632.             'postFile': '../AIlib2/weights/conf/channelEmergency/para.json',###后处理参数文件

  633.             'txtFontSize':40,###文本字符的大小

  634.             'digitFont': { 'line_thickness':2,'boxLine_thickness':1, 'fontSize':1.0,'waterLineColor':(0,255,255),'waterLineWidth':3},###显示框、线设置

  635.             'testImgPath':'../AIdemo2/images/channelEmergency/',###测试图像的位置

  636.             'testOutPath':'images/results/',###输出测试图像位置

  637.         }

  638.    

  639.     ###乡村路违法种植

  640.     if opt['business'] == 'countryRoad': 

  641.         par={

  642.             'device':'0', ###显卡号,如果用TRT模型，只支持0（单显卡）

  643.             'labelnames':"../AIlib2/weights/conf/countryRoad/labelnames.json", ###检测类别对照表

  644.             'max_workers':1, ###并行线程数

  645.             'trackPar':{'sort_max_age':2,'sort_min_hits':3,'sort_iou_thresh':0.2,'det_cnt':10,'windowsize':29,'patchCnt':100},

  646.             'postProcess':{'function':default_mix,'pars':{ }},

  647.             'models':

  648.                 [

  649.                     {

  650.                         'weight':"../weights/%s/AIlib2/%s/yolov5_%s_fp16.engine"%(opt['gpu'], opt['business'] ,opt['gpu']  ),###检测模型路径

  651.                         #'weight':'../AIlib2/weights/conf/%s/yolov5.pt'%(opt['business']  ),

  652.                         'name':'yolov5',

  653.                         'model':yolov5Model,

  654.                         'par':{ 'half':True,'device':'cuda:0' ,'conf_thres':0.25,'iou_thres':0.45,'allowedList':[0,1,2,3],'segRegionCnt':1, 'trtFlag_det':False,'trtFlag_seg':False, "score_byClass":{"0":0.25,"1":0.3,"2":0.3,"3":0.3  } },       

  655.                     }

  656.                 ],  

  657.             '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

  658.             'postFile': '../AIlib2/weights/conf/countryRoad/para.json',###后处理参数文件

  659.             'txtFontSize':40,###文本字符的大小

  660.             'digitFont': { 'line_thickness':2,'boxLine_thickness':1, 'fontSize':1.0,'waterLineColor':(0,255,255),'waterLineWidth':3},###显示框、线设置

  661.             'testImgPath':'../AIdemo2/images/countryRoad/',###测试图像的位置

  662.             'testOutPath':'images/results/',###输出测试图像位置

  663.         }

  664.    

  665.     ###城管项目，检测城市垃圾和车辆

  666.     if opt['business'] == 'cityMangement':

  667.         par={

  668.             'device':'0', ###显卡号,如果用TRT模型，只支持0（单显卡）

  669.             'labelnames':"../AIlib2/weights/conf/%s/labelnames.json"%(opt['business']), ###检测类别对照表

  670.             'max_workers':1, ###并行线程数

  671.             'trackPar':{'sort_max_age':2,'sort_min_hits':3,'sort_iou_thresh':0.2,'det_cnt':10,'windowsize':29,'patchCnt':100},

  672.             'postProcess':{'function':default_mix,'pars':{ }},

  673.             'models':

  674.                 [

  675.                     {

  676.                         'weight':"../weights/%s/AIlib2/%s/yolov5_%s_fp16.engine"%(opt['gpu'], opt['business'] ,opt['gpu']  ),###检测模型路径

  677.                         #'weight':'../AIlib2/weights/conf/%s/yolov5.pt'%(opt['business']  ),

  678.                         'name':'yolov5',

  679.                         'model':yolov5Model,

  680.                         'par':{ 'half':True,'device':'cuda:0' ,'conf_thres':0.25,'iou_thres':0.45,'allowedList':[0,1,2,3],'segRegionCnt':1, 'trtFlag_det':False,'trtFlag_seg':False, "score_byClass":{"0":0.25,"1":0.3,"2":0.3,"3":0.3  } },       

  681.                     }

  682.                 ],  

  683.             '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

  684.             'postFile': '../AIlib2/weights/conf/%s/para.json'%(opt['business']),###后处理参数文件

  685.             'txtFontSize':40,###文本字符的大小

  686.             'digitFont': { 'line_thickness':2,'boxLine_thickness':1, 'fontSize':1.0,'waterLineColor':(0,255,255),'waterLineWidth':3},###显示框、线设置

  687.             'testImgPath':'images/cityMangement',###测试图像的位置

  688.             'testOutPath':'images/results/',###输出测试图像位置

  689.         }

  690.     ###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)

  691.     ###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"    

  692.     if opt['business'] == 'cityRoad':

  693.         par={

  694.             'device':'0', ###显卡号,如果用TRT模型，只支持0（单显卡）

  695.             'labelnames':"../AIlib2/weights/conf/%s/labelnames.json"%(opt['business']), ###检测类别对照表

  696.             'max_workers':1, ###并行线程数

  697.             'trackPar':{'sort_max_age':2,'sort_min_hits':3,'sort_iou_thresh':0.2,'det_cnt':10,'windowsize':29,'patchCnt':100},

  698.             'postProcess':{'function':default_mix,'pars':{ }},

  699.             'models':

  700.                 [

  701.                     {

  702.                         'weight':"../weights/%s/AIlib2/%s/yolov5_%s_fp16.engine"%(opt['gpu'], opt['business'] ,opt['gpu']  ),###检测模型路径

  703.                         #'weight':'../AIlib2/weights/conf/%s/yolov5.pt'%(opt['business']  ),

  704.                         'name':'yolov5',

  705.                         'model':yolov5Model,

  706.                         'par':{ 'half':True,'device':'cuda:0' ,'conf_thres':0.25,'iou_thres':0.45,'allowedList':[0,1,2,3],'segRegionCnt':1, 'trtFlag_det':False,'trtFlag_seg':False, "score_byClass":{"0":0.25,"1":0.3,"2":0.3,"3":0.3  } },       

  707.                     }

  708.                 ],  

  709.             '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

  710.             'postFile': '../AIlib2/weights/conf/%s/para.json'%(opt['business']),###后处理参数文件

  711.             'txtFontSize':40,###文本字符的大小

  712.             'digitFont': { 'line_thickness':2,'boxLine_thickness':1, 'fontSize':1.0,'waterLineColor':(0,255,255),'waterLineWidth':3},###显示框、线设置

  713.             'testImgPath':'images/%s'%(opt['business']),###测试图像的位置

  714.             'testOutPath':'images/results/',###输出测试图像位置

  715.         }

  716.     if opt['business'] == 'illParking':

  717.         from utilsK.illParkingUtils import illParking_postprocess_N

  718.         par={

  719.             'device':'0', ###显卡号,如果用TRT模型，只支持0（单显卡）

  720.             'labelnames':"../AIlib2/weights/conf/%s/labelnames.json"%(opt['business']), ###检测类别对照表

  721.             'trackPar':{'sort_max_age':2,'sort_min_hits':3,'sort_iou_thresh':0.2,'det_cnt':10,'windowsize':29,'patchCnt':100},

  722.             'postProcess':{'function':illParking_postprocess_N, 'pars':{ }},

  723.             'models':

  724.                 [

  725.                     {

  726.                         'weight':"../weights/%s/AIlib2/%s/yolov5_%s_fp16.engine"%(opt['gpu'], opt['business'] ,opt['gpu']  ),###检测模型路径

  727.                         #'weight':'../AIlib2/weights/conf/%s/yolov5.pt'%(opt['business']  ),

  728.                         'name':'yolov5',

  729.                         'model':yolov5Model,

  730.                         'par':{ 'half':True,'device':'cuda:0' ,'conf_thres':0.25,'iou_thres':0.45,'allowedList':[0,1,2,3],'segRegionCnt':1, 'trtFlag_det':False,'trtFlag_seg':False, "score_byClass":{"0":0.25,"1":0.3,"2":0.3,"3":0.3  } },       

  731.                     }

  732.                 ],     

  733.             '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

  734.             'postFile': '../AIlib2/weights/conf/%s/para.json'%(opt['business'] ),###后处理参数文件

  735.             'txtFontSize':20,###文本字符的大小

  736.             'digitFont': { 'line_thickness':2,'boxLine_thickness':1, 'fontSize':1.0,'waterLineColor':(0,255,255),'waterLineWidth':2},###显示框、线设置

  737.             'testImgPath':'images/cityMangement',###测试图像的位置

  738.             'testOutPath':'images/results/',###输出测试图像位置

  739.         }

  740.         

  741.     #第一步加载模型

  742.     modelList=[ modelPar['model'](weights=modelPar['weight'],par=modelPar['par'])  for  modelPar in par['models'] ]

  743.     #第二步准备跟踪参数

  744.     trackPar=par['trackPar']

  745.     sort_tracker = Sort(max_age=trackPar['sort_max_age'],

  746.                        min_hits=trackPar['sort_min_hits'],

  747.                        iou_threshold=trackPar['sort_iou_thresh'])

  748.              

  749.     #第三步准备画图字体

  750.     drawPar = get_drawPar(par)

  751.  

  752.     #第四步获取图像测试及视频

  753.     imgpaths,videopaths = get_images_videos( par['testImgPath'])

  754. 

  755.     #第五步开始测试

  756.     for i in range(len(imgpaths)):

  757.         imgpath = imgpaths[i]

  758.         bname = os.path.basename(imgpath )

  759.         im0s=[cv2.imread(imgpath)]

  760.         time00 = time.time()

  761. 

  762.         retResults,timeOut = AI_det_track_batch_N(im0s, [i] ,modelList,par['postProcess'],sort_tracker ,trackPar)

  763.         '''

  764.         返回值retResults[2] --list,其中每一个元素为一个list，表示每一帧的检测结果，每一个结果是由多个list构成，每个list表示一个框，格式为[  x0 ,y0  ,x1  ,y1  ,conf,cls ,ifrmae,trackId ]

  765.                             --etc. retResults[2][j][k]表示第j帧的第k个框。

  766.         '''

  767.         time11 = time.time()

  768.         if len(retResults[1])>0:

  769.             retResults[0][0] = drawBoxTraceSimplied(retResults[1],i, retResults[0][0],rainbows=rainbows,boxFlag=True,traceFlag=False,names=drawPar['names'])

  770.         

  771.         image_array = retResults[0][0]

  772.         

  773.         cv2.imwrite(  os.path.join( par['testOutPath'],bname ) ,image_array )

  774.         print('----image:%s, Allprocess:%s  %s ,save:%s , objcnt:%d'%(bname,(time11-time00) * 1000, timeOut,(time.time() - time11) * 1000 ,len(retResults[2]))  )

  775.     ##第五步开始测试视频    

  776.     print('##begin to process videos, total %d videos'%( len(videopaths)))

  777.     for i,video in enumerate(videopaths):

  778.         print('process video%d :%s '%(i,video))

  779. 

  780.         par0={'modelList':modelList,'postProcess':par['postProcess'],'drawPar':drawPar,'outpth':par['testOutPath'], 'sort_tracker':sort_tracker,'trackPar':trackPar}

  781.         

  782.         process_video(video,par0,mode='track')      

  783. def OCR_demo2(opt):

  784.     from ocrUtils2 import crnn_model

  785.     from ocrUtils2.ocrUtils import get_cfg,recognition_ocr,strLabelConverter

  786.     

  787.     if opt['business'] == 'ocr2':

  788.          par={

  789.             'image_dir':'images/ocr_en',

  790.             'outtxt':'images/results',

  791.             'weights':'../AIlib2/weights/conf/ocr2/crnn_448X32.pth',

  792. 

  793.             #'weights':'../weights/2080Ti/AIlib2/ocr2/crnn_2080Ti_fp16_448X32.engine',

  794.             'device':'cuda:0',

  795.             'cfg':'../AIlib2/weights/conf/ocr2/360CC_config.yaml',

  796.             

  797.             'char_file':'../AIlib2/weights/conf/ocr2/chars.txt',

  798.             'imgH':32,

  799.             'imgW':448,

  800.             'workers':1

  801. 

  802.             

  803.         }  

  804.     image_dir=par['image_dir']

  805.     outtxt=par['outtxt']

  806.     workers=par['workers']

  807.     weights= par['weights']

  808.     device=par['device']

  809. 

  810.     char_file=par['char_file']

  811.     imgH=par['imgH']

  812.     imgW=par['imgW']

  813.     cfg = par['cfg']

  814.     

  815.    

  816.     

  817.     config = get_cfg(cfg, char_file)    

  818. 

  819. 

  820.     par['contextFlag']=False

  821.     device = torch.device('cuda:0') if torch.cuda.is_available() else torch.device('cpu') 

  822.     if weights.endswith('.pth'):

  823.         model = crnn_model.get_crnn(config,weights=weights).to(device)

  824.         par['model_mode']='pth'

  825.     else:    

  826.         logger = trt.Logger(trt.Logger.ERROR)

  827.         with open(weights, "rb") as f, trt.Runtime(logger) as runtime:

  828.             model = runtime.deserialize_cuda_engine(f.read())# 输入trt本地文件，返回ICudaEngine对象

  829.             print('#####load TRT file:',weights,'success #####')

  830.         context = model.create_execution_context() 

  831.         par['model_mode']='trt';par['contextFlag']=context

  832. 

  833.     converter = strLabelConverter(config.DATASET.ALPHABETS)

  834. 

  835.     img_urls=glob.glob('%s/*.jpg'%( image_dir ))

  836.     img_urls.extend( glob.glob('%s/*.png'%( image_dir ))  )

  837.     cnt=len(img_urls)

  838.     print('%s  has %d images'%(image_dir ,len(img_urls) )  )

  839.     # 准备数据

  840.     parList=[]

  841.     for i in range(cnt):

  842.         img_patch=cv2.imread( img_urls[i] ,   cv2.IMREAD_GRAYSCALE) 

  843.         started = time.time()

  844.         img = cv2.imread(img_urls[i])

  845.         sim_pred = recognition_ocr(config, img, model, converter, device,par=par)

  846.         finished = time.time()

  847.         print('{0}: elapsed time: {1} prd:{2} '.format( os.path.basename( img_urls[i] ), finished - started, sim_pred ))

  848.         

  849. def customization_demo(opt):

  850.     from AI import AI_process_C     

  851.         

  852.     if opt['business'] == 'channel2': 

  853.         

  854.         from ocr import ocrModel

  855.         

  856.         par={

  857.                 'device':'0', ###显卡号,如果用TRT模型，只支持0（单显卡）

  858.                 'labelnames':"../AIlib2/weights/conf/%s/labelnames.json"%(opt['business']), ###检测类别对照表

  859.                 'max_workers':1, ###并行线程数

  860.                 'postProcess':{'function':default_mix,'pars':{'objs':[2] },'name':'channel2',},     

  861.                 'models':[

  862.                         {

  863.                             #'weight':"../weights/%s/AIlib2/%s/yolov5_%s_fp16.engine"%(opt['gpu'], opt['business'] ,opt['gpu']  ),###检测模型路径

  864.                             'weight':'../AIlib2/weights/conf/%s/yolov5.pt'%(opt['business']  ),

  865.                             #'weight':'/mnt/thsw2/DSP2/weights/channel2/yolov5.pt',

  866.                             #'weight':'/mnt/thsw2/DSP2/weights/channel2/yolov5_2080Ti_fp16.engine',

  867.                             'name':'yolov5',

  868.                             'model':yolov5Model,

  869.                             'par':{ 'half':True,'device':'cuda:0' ,'conf_thres':0.1,'iou_thres':0.45,'allowedList':list(range(20)),'segRegionCnt':1, 'trtFlag_det':False,'trtFlag_seg':False, "score_byClass":{str(x):0.1 for x in range(20) } }

  870.                         

  871.                         },

  872.                       

  873.                         {                      

  874.                             'weight' : '../weights/%s/AIlib2/ocr2/crnn_ch_%s_fp16_192X32.engine'%(opt['gpu'], opt['gpu']),  

  875.                             'name':'ocr',

  876.                             'model':ocrModel,

  877.                             'par':{

  878.                                 'char_file':'../AIlib2/weights/conf/ocr2/chars.txt',

  879.                                 'mode':'ch',

  880.                                 'nc':3,

  881.                                 'imgH':32,

  882.                                 'imgW':192,

  883.                                 'hidden':256,

  884.                                 'mean':[0.5,0.5,0.5],

  885.                                 'std':[0.5,0.5,0.5],

  886.                                 'dynamic':False,

  887.                                 },

  888.                         }  

  889.                     ],

  890.                 

  891.                 '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

  892.                 'postFile': '../AIlib2/weights/conf/%s/para.json'%(opt['business'] ),###后处理参数文件

  893.                 'txtFontSize':20,###文本字符的大小

  894.                 'digitFont': { 'line_thickness':2,'boxLine_thickness':1, 'fontSize':1.0,'waterLineColor':(0,255,255),'segLineShow':True,'waterLineWidth':2},###显示框、线设置

  895.                 'testImgPath':'images/%s/'%(opt['business']),

  896.                 #'testImgPath':'/home/thsw2/WJ/src/OCR/shipNames',

  897.                 'testOutPath':'images/results/',###输出测试图像位置    

  898.             }

  899.             

  900.     if opt['business'] == 'crackMeasurement': 

  901.         from utilsK.crackUtils import Crack_measure

  902.         print(  '%s 只能测试图像，不能测试视频%s'%('#'*20,'#'*20))

  903.         par={

  904.                 'device':'0', ###显卡号,如果用TRT模型，只支持0（单显卡）

  905.                 'labelnames':"../AIlib2/weights/conf/%s/labelnames.json"%(opt['business']), ###检测类别对照表

  906.                 'max_workers':1, ###并行线程数

  907.                 'postProcess':{

  908.                     'name':'crackMeasurement',

  909.                     'function':Crack_measure,

  910.                     'pars':{'dsx':(123-30)*1000/35*0.004387636  ,'objs':[0,1,2]}

  911.                     }, 

  912.                 'models':[

  913.                         {

  914.                             #'weight':"../weights/%s/AIlib2/%s/yolov5_%s_fp16.engine"%(opt['gpu'], opt['business'] ,opt['gpu']  ),###检测模型路径

  915.                             'weight':'../AIlib2/weights/conf/%s/yolov5.pt'%(opt['business']  ),

  916.                             'name':'yolov5',

  917.                             'model':yolov5Model,

  918.                             'par':{ 'half':True,'device':'cuda:0' ,'conf_thres':0.1,'iou_thres':0.45,'allowedList':[0,1,2,3],'segRegionCnt':1, 'trtFlag_det':False,'trtFlag_seg':False, "score_byClass":{"0":0.1,"1":0.1,"2":0.1 } }

  919.                         

  920.                         },

  921.                       

  922.                         {                      

  923.                             #'weight':"../weights/%s/AIlib2/%s/stdc_360X640_%s_fp16.engine"%(opt['gpu'], opt['business'] ,opt['gpu']  ),###检测模型路径

  924.                             'weight':'../AIlib2/weights/conf/%s/stdc_360X640.pth'%(opt['business']  ),

  925.                             'par':{

  926.                                 #'modelSize':(640,360),

  927.                                 'modelSize':(1920,1080),

  928.                                 'dynamic':True,

  929.                                 'mean':(0.485, 0.456, 0.406),'std' :(0.229, 0.224, 0.225),'predResize':True,'numpy':False, 'RGB_convert_first':True,'seg_nclass':2},###分割模型预处理参数

  930.                             'model':stdcModel,

  931.                             'name':'stdc'        

  932.                             

  933.                         }

  934.                         

  935.               

  936.                        

  937.                     ],

  938.                 

  939.                 '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

  940.                 'postFile': '../AIlib2/weights/conf/%s/para.json'%(opt['business'] ),###后处理参数文件

  941.                 'txtFontSize':20,###文本字符的大小

  942.                 'digitFont': { 'line_thickness':2,'boxLine_thickness':1, 'fontSize':1.0,'waterLineColor':(0,255,255),'segLineShow':True,'waterLineWidth':2},###显示框、线设置

  943.                 'testImgPath':'images/%s/'%(opt['business']),

  944.                 #'testImgPath':'/mnt/thsw2/DSP2/weights/cityMangement2_1102/images/debug',

  945.                 'testOutPath':'images/results/',###输出测试图像位置    

  946.             }

  947.                

  948.     #第一步加载模型

  949.     modelList=[ modelPar['model'](weights=modelPar['weight'],par=modelPar['par'])  for  modelPar in par['models'] ]

  950.     print(' load moder over')

  951. 

  952.     #准备画图字体

  953.     labelnames = par['labelnames'] ##对应类别表

  954.     names=get_labelnames(labelnames)

  955.     label_arraylist = get_label_arrays(names,rainbows,outfontsize=par['txtFontSize'],fontpath="../AIlib2/conf/platech.ttf")

  956. 

  957.     #图像测试

  958.     imgpaths,videopaths = get_images_videos( par['testImgPath'])

  959. 

  960. 

  961.     #开始测试

  962.     for imgUrl in   imgpaths[0:]:

  963.         img = cv2.imread(imgUrl);bname = os.path.basename(imgUrl)

  964.         if opt['business'] == 'crackMeasurement':

  965.             ret,timeInfos = AI_process_C([img],modelList,par['postProcess'])

  966.             #返回类型ret-list,[[ x0,y0,x1,y1,score,class,裂缝长度,平均宽度,最大宽度,最小宽度],[...],[...]]

  967.             for re in ret:

  968.                 print('Summarized Cracklength = %.1f mm Mean crack width = %.1f mm  Max crack width = %.1f mm Min crack width = %.1f mm '%( re[6], re[7],re[8],re[9] ) )

  969.         elif opt['business'] == 'channel2':

  970.             ret,timeInfos = AI_process_C([img],modelList,par['postProcess'])

  971.         

  972.         timeInfos=bname+':'+timeInfos

  973.         print(timeInfos,ret )

  974.         if len(ret)>0:

  975.             img0 = drawAllBox(ret,img,label_arraylist,rainbows,par['digitFont'])

  976.         else:  img0=    img

  977.         cv2.imwrite(os.path.join('images/results/',bname ) ,img0)

  978. 

  979.         

  980.         

  981.     

  982.     

  983.                 

  984. def OBB_track_demo(opt):

  985.     ###倾斜框（OBB）的ship目标检测

  986.     par={

  987.      

  988.         'obbModelPar':{

  989.             'model_size':(608,608),'K':100,'conf_thresh':0.3, 'down_ratio':4,'num_classes':15,'dataset':'dota',

  990.             'heads': {'hm': None,'wh': 10,'reg': 2,'cls_theta': 1},

  991.             'mean':(0.5, 0.5, 0.5),'std':(1, 1, 1), 'half': False,'decoder':None,

  992.             'weights':'../weights/%s/AIlib2/%s/obb_608X608_%s_fp16.engine'%(opt['gpu'],opt['business'],opt['gpu']),

  993.             

  994.         },

  995.         'outpth': 'images/results',

  996.         'trackPar':{'sort_max_age':2,'sort_min_hits':3,'sort_iou_thresh':0.2,'det_cnt':10,'windowsize':29,'patchCnt':100},

  997.         'device':"cuda:0",

  998.         

  999.         #'test_dir': '/mnt/thsw2/DSP2/videos/obbShips/DJI_20230208110806_0001_W_6M.MP4',

  1000.         'test_dir':'/mnt/thsw2/DSP2/videos/obbShips/freighter2.mp4',

  1001.         

  1002.         'test_flag':True,

  1003.         'postFile': '../AIlib2/weights/conf/%s/para.json'%(opt['business'] ),###后处理参数文件

  1004.         'drawBox':True,#####是否画框

  1005.         'drawPar': { 'digitWordFont' :{'line_thickness':2,'boxLine_thickness':1,'wordSize':40, 'fontSize':1.0,'label_location':'leftTop'}} ,  

  1006.         'labelnames':"../AIlib2/weights/conf/%s/labelnames.json"%(opt['business'] ), ###检测类别对照表

  1007.     }

  1008.     #par['model_size'],par['mean'],par['std'],par['half'],par['saveType'],par['heads'],par['labelnames'],par['decoder'],par['down_ratio'],par['drawBox']

  1009.     #par['rainbows'],par['label_array'],par['digitWordFont']

  1010. 

  1011.     obbModelPar = par['obbModelPar']

  1012.     ####加载模型

  1013.     model,decoder2=load_model_decoder_OBB(obbModelPar)

  1014.     obbModelPar['decoder']=decoder2

  1015. 

  1016.     names=get_labelnames(par['labelnames']);obbModelPar['labelnames']=names

  1017.     

  1018.     _,_,_,rainbows=get_postProcess_para(par['postFile']);par['drawPar']['rainbows']=rainbows

  1019. 

  1020.     

  1021.     label_arraylist = get_label_arrays(names,rainbows,outfontsize=par['drawPar']['digitWordFont']['wordSize'],fontpath="../AIlib2/conf/platech.ttf")

  1022.     #par['label_array']=label_arraylist

  1023. 

  1024.     trackPar=par['trackPar']

  1025.     sort_tracker = OBB_Sort(max_age=trackPar['sort_max_age'],

  1026.                        min_hits=trackPar['sort_min_hits'],

  1027.                        iou_threshold=trackPar['sort_iou_thresh'])

  1028.                        

  1029.     

  1030.      ##图像测试和视频

  1031.     impth = par['test_dir']

  1032.     img_urls=[]###获取文件里所有的图像

  1033.     video_urls=[]###获取文件里所有的视频

  1034.     img_postfixs = ['.jpg','.JPG','.PNG','.png'];

  1035.     vides_postfixs= ['.MP4','.mp4','.avi']

  1036.     if os.path.isdir(impth):

  1037.         for postfix in img_postfixs:

  1038.             img_urls.extend(glob.glob('%s/*%s'%(impth,postfix ))  )

  1039.         for postfix in ['.MP4','.mp4','.avi']:

  1040.             video_urls.extend(glob.glob('%s/*%s'%(impth,postfix ))  )

  1041.     else:

  1042.         postfix = os.path.splitext(impth)[-1]

  1043.         if postfix in img_postfixs: img_urls=[ impth ]

  1044.         if postfix in vides_postfixs: video_urls = [impth ]

  1045.         

  1046.     parIn = {'obbModelPar':obbModelPar,'modelPar':{'obbmodel': model},'sort_tracker':sort_tracker,'outpth':par['outpth'],'trackPar':trackPar,'drawPar':par['drawPar']}

  1047.     par['drawPar']['label_array']=label_arraylist        

  1048.     for img_url in img_urls:

  1049.         #print(img_url)

  1050.         ori_image=cv2.imread(img_url)     

  1051.         

  1052.         #ori_image_list,infos = OBB_infer(model,ori_image,obbModelPar)         

  1053.         

  1054.         ori_image_list,infos  = OBB_tracker_batch([ori_image],[0],parIn['modelPar'],parIn['obbModelPar'],None,parIn['trackPar'],None)

  1055.      

  1056.         ori_image_list[1] =  draw_obb(ori_image_list[2] ,ori_image_list[1],par['drawPar'])

  1057.         

  1058.         imgName = os.path.basename(img_url)

  1059.         saveFile = os.path.join(par['outpth'], imgName)

  1060.         ret=cv2.imwrite(saveFile, ori_image_list[1])

  1061.         if not ret:

  1062.             print(saveFile, ' not created ')

  1063.         print(  os.path.basename(img_url),':',infos,ori_image_list[2])

  1064.         

  1065.     ###处理视频

  1066.     

  1067.     for video_url in video_urls:

  1068.         process_video(video_url, parIn ,mode='obbTrack')

  1069. 

  1070. def crowd_demo(opt):

  1071.     if opt['business']=='crowdCounting':

  1072.      

  1073.         from crowd import crowdModel as Model

  1074.         par={

  1075.             'mean':[0.485, 0.456, 0.406], 'std':[0.229, 0.224, 0.225],'threshold':0.5, 

  1076.             'input_profile_shapes':[(1,3,256,256),(1,3,1024,1024),(1,3,2048,2048)],

  1077.             'modelPar':{'backbone':'vgg16_bn', 'gpu_id':0,'anchorFlag':False, 'width':None,'height':None ,'line':2, 'row':2},

  1078. 

  1079.             'weights':"../weights/%s/AIlib2/%s/crowdCounting_%s_dynamic.engine"%(opt['gpu'], opt['business'] ,opt['gpu']  ),###检测模型路径

  1080.             'testImgPath':'images/%s'%(opt['business'] ),###测试图像的位置

  1081.             'testOutPath':'images/results/',###输出测试图像位置

  1082.             

  1083.             }

  1084.      

  1085.     

  1086.     #weights='weights/best_mae.pth'

  1087.     cmodel = Model(par['weights'],par)

  1088. 

  1089.     img_path = par['testImgPath']

  1090.     File = os.listdir(img_path)

  1091.     targetList = []

  1092.     for file in File[0:]:

  1093.         COORlist = []

  1094.         imgPath = img_path + os.sep + file

  1095.         

  1096.         

  1097.         img_raw = cv2.cvtColor(cv2.imread(imgPath),cv2.COLOR_BGR2RGB)

  1098.         # cmodel.eval---

  1099.         # 输入读取的RGB数组

  1100.         # 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：

  1101.         # [ [ [ (x0,y0),(x1,y1),(x2,y2),(x3,y3)  ],score， cls ],  [ [ (x0,y0),(x1,y1),(x2,y2),(x3,y3)  ],score ，cls ],........  ]

  1102.         

  1103.         prets, infos = cmodel.eval(img_raw)

  1104.         

  1105.         

  1106.         print(file,infos,' 人数：',len(prets[1]))

  1107.         

  1108.  

  1109.         img_to_draw = cv2.cvtColor(np.array(img_raw), cv2.COLOR_RGB2BGR)

  1110.         # 打印预测图像中人头的个数

  1111.         for p in prets[1]:

  1112.             img_to_draw = cv2.circle(img_to_draw, (int(p[0]), int(p[1])), 2, (0, 255, 0), -1)

  1113.             COORlist.append((int(p[0]), int(p[1])))

  1114.         # 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),..], ...]

  1115.         targetList.append(COORlist)

  1116.         #time.sleep(2)

  1117.         # 保存预测图片

  1118.         cv2.imwrite(os.path.join(par['testOutPath'], file), img_to_draw)

  1119.  

  1120. if __name__=="__main__":

  1121. 

  1122.     #jkm_demo()

  1123.     businessAll=['river', 'river2','highWay2','noParking','drowning','forest2','vehicle','pedestrian','smogfire' , 'AnglerSwimmer','channelEmergency', 'countryRoad','cityMangement','ship2','cityMangement2','cityRoad','illParking',"crowdCounting",'cityMangement3','crackMeasurement','pothole','channel2','riverT']

  1124.     businessAll = [ 'forest2'     ]

  1125.     

  1126.     for busi in businessAll:

  1127.         print('-'*40,'beg to test:',busi,'-'*40)

  1128.         opt={'gpu':'2080Ti','business':busi}

  1129.         if busi in ['ship2']:

  1130.             OBB_track_demo(opt)

  1131.         elif  opt['business'] in ['crowdCounting'] :

  1132.             crowd_demo(opt)   

  1133.         elif opt['business'] in ['crackMeasurement','channel2'] :

  1134.             customization_demo(opt)

  1135.         else:

  1136.         #if opt['business'] in ['river','highWay2','noParking','drowning','']:

  1137.             det_track_demo_N(opt ) 

  1138.             

  1139.   

  1140.