wangjin0928
/
AIdemo2
Watch 1
Star 0
Fork 0
Code Issues 0 Pull Requests 0 Releases 0 Wiki Activity
You can not select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.
48 Commits
1 Branch
Tree: ae7f97481a
Branches Tags
${ item.name }
Create branch ${ searchTerm }
from 'ae7f97481a'
${ noResults }
AIdemo2/demo3.0.py

969 lines
57KB
Raw Blame History 

  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

  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 get_drawPar(par):

  33.     labelnames = par['labelnames']

  34.     names=get_labelnames(labelnames)

  35.     mode_paras=par['detModelpara']

  36.     allowedList,allowedList_string=get_needed_objectsIndex(mode_paras) 

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

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

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

  40.     return drawPar

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

  42.     rate = num / total

  43.     time_n=time.time()

  44.     rate_num = int(rate * 30)

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

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

  47.     sys.stdout.write(r)

  48.     sys.stdout.flush()

  49.         

  50. 

  51. '''

  52.     多线程

  53. '''

  54. 

  55. def process_v1(frame):

  56.     #try:

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

  58.     time00 = time.time()

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

  60.     

  61.     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])

  62.      

  63.     time11 = time.time()

  64.     image_array = p_result[1]

  65.     

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

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

  68.     if len(p_result)==5:

  69.         image_mask = p_result[4]

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

  71.     

  72.     boxes=p_result[2]

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

  74.         for box in boxes:

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

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

  77.             fp.write(out_str)

  78.     time22 = time.time()

  79.     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)

  80.     return 'success'

  81.     #except Exception as e:

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

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

  84.     cap=cv2.VideoCapture(video)

  85.     if not  cap.isOpened():

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

  87.         return False  

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

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

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

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

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

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

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

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

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

  97.     num=0

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

  99.     

  100.     imgarray_list = []; iframe_list = []

  101.     

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

  103.     ##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。

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

  105. 

  106. 

  107.     ##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.

  108.     while cap.isOpened():

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

  110.         iframe +=1

  111.         if not ret:break

  112.         if mode=='detSeg':

  113.             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'])

  114.         elif mode == 'track':

  115.             #sampleCount=10

  116.             imgarray_list.append( imgarray )

  117.             iframe_list.append(iframe )    

  118.             if iframe%patch_cnt==0:

  119.                 time_patch0 = time.time()

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

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

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

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

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

  125.                 time_patch2 = time.time()                    

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

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

  128.                     img_draw = imgarray_list[ iiframe- frame_min  ]

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

  130.                     ret = vid_writer_AI.write(img_draw)        

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

  132.                 imgarray_list=[];iframe_list=[]     

  133.         elif mode =='obbTrack':

  134.             imgarray_list.append( imgarray )

  135.             iframe_list.append(iframe )    

  136.             if iframe%patch_cnt==0:

  137.                 time_patch0 = time.time()

  138.                     

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

  140.                 print(  timeInfos )

  141.                 

  142.                 #对结果画图

  143.                 track_det_np = track_det_results[1]

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

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

  146.                     img_draw = imgarray_list[ iiframe- frame_min  ]

  147.                     

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

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

  150.                         if  True: 

  151.                             frameIdex=12;trackIdex=13;

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

  153.                             

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

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

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

  157.                             for trackId in track_ids:

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

  159.                                 xcs = boxes_before_oneFrame_oneId[:,8]

  160.                                 ycs = boxes_before_oneFrame_oneId[:,9]

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

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

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

  164. 

  165.                     ret = vid_writer_AI.write(img_draw)   

  166.                    

  167.                 #sys.exit(0)

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

  169.                 imgarray_list=[];iframe_list=[]     

  170.   

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

  172.             

  173.         else:

  174.             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'])

  175.             

  176. 

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

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

  179.             ret = vid_writer_AI.write(image_array)    

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

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

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

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

  184.                 post_results=[]   

  185.     

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

  187.                 post_results.append(p_result)

  188.  

  189.     vid_writer_AI.release();       

  190.    

  191. def det_track_demo_N(business ):

  192.     ####河道巡检的跟踪模型参数

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

  194.         from utilsK.queRiver import riverDetSegMixProcess_N

  195.         par={

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

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

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

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

  200.             '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         

  201.             'models':

  202.                 [

  203.                     {

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

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

  206.                         'name':'yolov5',

  207.                         'model':yolov5Model,

  208.                         '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  } },       

  209.                     },

  210.                     {

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

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

  213.                         'par':{

  214.                             '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

  215.                         'model':stdcModel,

  216.                         'name':'stdc'        

  217.                     }

  218.                         

  219.                 ],

  220.             '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

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

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

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

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

  225.             'testImgPath':'images/river2',###测试图像的位置

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

  227.             

  228.         }

  229. 

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

  231.         from  segutils.trafficUtils  import tracfficAccidentMixFunction_N 

  232.         par={

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

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

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

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

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

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

  239.                 },

  240.             'models':

  241.                 [

  242.                     {

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

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

  245.                         'name':'yolov5',

  246.                         'model':yolov5Model,

  247.                         '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  } },       

  248.                     },

  249.                     {

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

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

  252.                         'par':{

  253.                             '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

  254.                         'model':stdcModel,

  255.                         'name':'stdc'        

  256.                     }

  257.                         

  258.                 ],

  259.                 

  260. 

  261.             '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

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

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

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

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

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

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

  268.         }

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

  270.           

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

  272.         from utilsK.noParkingUtils import mixNoParking_road_postprocess_N

  273.         par={

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

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

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

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

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

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

  280.                         } ,

  281.             'models':

  282.                 [

  283.                     {

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

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

  286.                         'name':'yolov5',

  287.                         'model':yolov5Model,

  288.                         '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  } },       

  289.                     },

  290.                     {

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

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

  293.                         'par':{

  294.                             '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

  295.                         'model':stdcModel,

  296.                         'name':'stdc'        

  297.                     }

  298.                         

  299.                 ],

  300.             '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

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

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

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

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

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

  306.         }

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

  308.         

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

  310.         from utilsK.drownUtils import mixDrowing_water_postprocess_N

  311.         par={

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

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

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

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

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

  317.                         'pars':{ }},

  318.             'models':

  319.                 [

  320.                     {

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

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

  323.                         'name':'yolov5',

  324.                         'model':yolov5Model,

  325.                         '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  } },       

  326.                     },

  327.                     {

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

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

  330.                         'par':{

  331.                             '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

  332.                         'model':stdcModel,

  333.                         'name':'stdc'        

  334.                     }

  335.                         

  336.                 ],

  337.             '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    

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

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

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

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

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

  343.         }

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

  345.         

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

  347.         from DMPR import DMPRModel

  348.         from DMPRUtils.jointUtil import  dmpr_yolo_stdc

  349. 

  350.         par={

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

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

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

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

  355.             'postProcess':{

  356.                     'function':dmpr_yolo_stdc,

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

  358.                     }, 

  359.             'models':

  360.                 [

  361.                     {

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

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

  364.                         'name':'yolov5',

  365.                         'model':yolov5Model,

  366.                         '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  } }

  367.                     

  368.                     },

  369.                     {

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

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

  372.                         'par':{

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

  374.                             'name':'dmpr'

  375.                              },

  376.                         'model':DMPRModel,     

  377.                         'name':'dmpr'

  378.                     }

  379.             ]   ,

  380.             '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

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

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

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

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

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

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

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

  388.         }

  389.         

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

  391.         from DMPR import DMPRModel

  392.         from DMPRUtils.jointUtil import  dmpr_yolo_stdc

  393.         par={

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

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

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

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

  398.                 'postProcess':{

  399.                     'function':dmpr_yolo_stdc,

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

  401.                     }, 

  402.                 'models':[

  403.                         {

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

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

  406.                             'name':'yolov5',

  407.                             'model':yolov5Model,

  408.                             '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  } }

  409.                         

  410.                         },

  411.                         {

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

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

  414.                             'par':{

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

  416.                                 'name':'dmpr'

  417.                                  },

  418.                             'model':DMPRModel,     

  419.                             'name':'dmpr'

  420.                         },

  421.                         {

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

  423.                           

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

  425.                             'par':{

  426.                                 '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

  427.                             'model':stdcModel,

  428.                             'name':'stdc'        

  429.                             

  430.                         } 

  431.                     ],

  432.                 

  433.                 '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

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

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

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

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

  438.                 'testImgPath':'/mnt/thsw2/DSP2/weights/cityMangement3/images/debug',

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

  440.             }

  441.     

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

  443.         

  444.         par={

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

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

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

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

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

  450.             'models':

  451.                 [

  452.                     {

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

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

  455.                         'name':'yolov5',

  456.                         'model':yolov5Model,

  457.                         '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  } },       

  458.                     }

  459.                     

  460.                         

  461.                 ],

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

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

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

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

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

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

  468.         }

  469.      ###车辆巡检参数

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

  471.         par={

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

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

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

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

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

  477.             'models':

  478.                 [

  479.                     {

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

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

  482.                         'name':'yolov5',

  483.                         'model':yolov5Model,

  484.                         '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  } },       

  485.                     }

  486.                     

  487.                         

  488.                 ], 

  489.             '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

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

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

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

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

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

  495.         }    

  496.      ###行人检测模型

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

  498.         par={

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

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

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

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

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

  504.             'models':

  505.                 [

  506.                     {

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

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

  509.                         'name':'yolov5',

  510.                         'model':yolov5Model,

  511.                         '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  } },       

  512.                     }

  513.                     

  514.                         

  515.                 ], 

  516.           

  517.             '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

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

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

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

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

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

  523.         }

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

  525.         par={

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

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

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

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

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

  531.             'models':

  532.                 [

  533.                     {

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

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

  536.                         'name':'yolov5',

  537.                         'model':yolov5Model,

  538.                         '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  } },       

  539.                     }

  540.                 ],  

  541.             '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

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

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

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

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

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

  547.         }

  548.     ###钓鱼游泳检测  

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

  550.         par={

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

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

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

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

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

  556.             'models':

  557.                 [

  558.                     {

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

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

  561.                         'name':'yolov5',

  562.                         'model':yolov5Model,

  563.                         '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  } },       

  564.                     }

  565.                 ],  

  566.             '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

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

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

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

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

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

  572.         }

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

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

  575.         par={

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

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

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

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

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

  581.             'models':

  582.                 [

  583.                     {

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

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

  586.                         'name':'yolov5',

  587.                         'model':yolov5Model,

  588.                         '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  } },       

  589.                     }

  590.                 ],  

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

  592.             

  593.            

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

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

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

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

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

  599.         }

  600.    

  601.     ###乡村路违法种植

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

  603.         par={

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

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

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

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

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

  609.             'models':

  610.                 [

  611.                     {

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

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

  614.                         'name':'yolov5',

  615.                         'model':yolov5Model,

  616.                         '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  } },       

  617.                     }

  618.                 ],  

  619.             '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

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

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

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

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

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

  625.         }

  626.    

  627.     ###城管项目，检测城市垃圾和车辆

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

  629.         par={

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

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

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

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

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

  635.             'models':

  636.                 [

  637.                     {

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

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

  640.                         'name':'yolov5',

  641.                         'model':yolov5Model,

  642.                         '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  } },       

  643.                     }

  644.                 ],  

  645.             '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

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

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

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

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

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

  651.         }

  652.     ###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)

  653.     ###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"    

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

  655.         par={

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

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

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

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

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

  661.             'models':

  662.                 [

  663.                     {

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

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

  666.                         'name':'yolov5',

  667.                         'model':yolov5Model,

  668.                         '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  } },       

  669.                     }

  670.                 ],  

  671.             '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

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

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

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

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

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

  677.         }

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

  679.         from utilsK.illParkingUtils import illParking_postprocess_N

  680.         par={

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

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

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

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

  685.             'models':

  686.                 [

  687.                     {

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

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

  690.                         'name':'yolov5',

  691.                         'model':yolov5Model,

  692.                         '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  } },       

  693.                     }

  694.                 ],     

  695.             '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

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

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

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

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

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

  701.         }

  702.         

  703.     #第一步加载模型

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

  705.     #第二步准备跟踪参数

  706.     trackPar=par['trackPar']

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

  708.                        min_hits=trackPar['sort_min_hits'],

  709.                        iou_threshold=trackPar['sort_iou_thresh'])

  710.              

  711.     #第三步准备画图字体

  712.     drawPar = get_drawPar(par)

  713.  

  714.     #第四步获取图像测试及视频

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

  716. 

  717.     #第五步开始测试

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

  719.         imgpath = imgpaths[i]

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

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

  722.         time00 = time.time()

  723. 

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

  725.         time11 = time.time()

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

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

  728.         

  729.         image_array = retResults[0][0]

  730.         '''

  731.         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 ]

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

  733.         '''

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

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

  736.     ##第五步开始测试视频    

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

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

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

  740. 

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

  742.         

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

  744. def OCR_demo2(opt):

  745.     from ocrUtils2 import crnn_model

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

  747.     

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

  749.          par={

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

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

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

  753. 

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

  755.             'device':'cuda:0',

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

  757.             

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

  759.             'imgH':32,

  760.             'imgW':448,

  761.             'workers':1

  762. 

  763.             

  764.         }  

  765.     image_dir=par['image_dir']

  766.     outtxt=par['outtxt']

  767.     workers=par['workers']

  768.     weights= par['weights']

  769.     device=par['device']

  770. 

  771.     char_file=par['char_file']

  772.     imgH=par['imgH']

  773.     imgW=par['imgW']

  774.     cfg = par['cfg']

  775.     

  776.    

  777.     

  778.     config = get_cfg(cfg, char_file)    

  779. 

  780. 

  781.     par['contextFlag']=False

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

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

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

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

  786.     else:    

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

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

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

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

  791.         context = model.create_execution_context() 

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

  793. 

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

  795. 

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

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

  798.     cnt=len(img_urls)

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

  800.     # 准备数据

  801.     parList=[]

  802.     for i in range(cnt):

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

  804.         started = time.time()

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

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

  807.         finished = time.time()

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

  809.         

  810. 

  811.     

  812.     

  813.                 

  814. def OBB_track_demo(opt):

  815.     ###倾斜框（OBB）的ship目标检测

  816.     par={

  817.      

  818.         'obbModelPar':{

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

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

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

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

  823.             

  824.         },

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

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

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

  828.         

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

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

  831.         

  832.         'test_flag':True,

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

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

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

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

  837.     }

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

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

  840. 

  841.     obbModelPar = par['obbModelPar']

  842.     ####加载模型

  843.     model,decoder2=load_model_decoder_OBB(obbModelPar)

  844.     obbModelPar['decoder']=decoder2

  845. 

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

  847.     

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

  849. 

  850.     

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

  852.     #par['label_array']=label_arraylist

  853. 

  854.     trackPar=par['trackPar']

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

  856.                        min_hits=trackPar['sort_min_hits'],

  857.                        iou_threshold=trackPar['sort_iou_thresh'])

  858.                        

  859.     

  860.      ##图像测试和视频

  861.     impth = par['test_dir']

  862.     img_urls=[]###获取文件里所有的图像

  863.     video_urls=[]###获取文件里所有的视频

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

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

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

  867.         for postfix in img_postfixs:

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

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

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

  871.     else:

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

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

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

  875.         

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

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

  878.     for img_url in img_urls:

  879.         #print(img_url)

  880.         ori_image=cv2.imread(img_url)     

  881.         

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

  883.         

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

  885.      

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

  887.         

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

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

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

  891.         if not ret:

  892.             print(saveFile, ' not created ')

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

  894.         

  895.     ###处理视频

  896.     

  897.     for video_url in video_urls:

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

  899. 

  900. def crowd_demo(opt):

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

  902.      

  903.         from crowd import crowdModel as Model

  904.         par={

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

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

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

  908. 

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

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

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

  912.             

  913.             }

  914.      

  915.     

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

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

  918. 

  919.     img_path = par['testImgPath']

  920.     File = os.listdir(img_path)

  921.     targetList = []

  922.     for file in File[0:]:

  923.         COORlist = []

  924.         imgPath = img_path + os.sep + file

  925.         

  926.         

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

  928.         # cmodel.eval---

  929.         # 输入读取的RGB数组

  930.         # 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：

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

  932.         

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

  934.         

  935.         

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

  937.         

  938.  

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

  940.         # 打印预测图像中人头的个数

  941.         for p in prets[1]:

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

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

  944.         # 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),..], ...]

  945.         targetList.append(COORlist)

  946.         #time.sleep(2)

  947.         # 保存预测图片

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

  949.  

  950. if __name__=="__main__":

  951. 

  952.     #jkm_demo()

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

  954.     businessAll = [ 'cityMangement3'     ]

  955.     

  956.     for busi in businessAll:

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

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

  959.         if busi in ['ship2']:

  960.             OBB_track_demo(opt)

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

  962.             crowd_demo(opt)    

  963.         else:

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

  965.             det_track_demo_N(opt ) 

  966.             

  967.   

  968.