gitadmin
/
tuoheng_algN
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

1)新增M027:建筑物下行人检测及计数 2)密集人群计数模型及自研车牌模型优化 3)分类模型支持pt模型加载（巴中水利）

This commit is contained in:
th 2025-07-25 18:20:18 +08:00
parent 12a4b296e1
commit d369031085
6 changed files with 747 additions and 319 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

203
concurrency/FileUploadThread.py
View File

@ -8,32 +8,32 @@ from loguru import logger
import cv2
from entity.FeedBack import message_feedback
from enums.ExceptionEnum import ExceptionType
from enums.ModelTypeEnum import ModelType
from exception.CustomerException import ServiceException
from util.AliyunSdk import AliyunOssSdk
from util.MinioSdk import MinioSdk
from util import TimeUtils
from enums.AnalysisStatusEnum import AnalysisStatus
from util.PlotsUtils import draw_painting_joint
from util.PlotsUtils import draw_painting_joint, draw_name_ocr, draw_name_crowd
from util.QueUtil import put_queue, get_no_block_queue, clear_queue
import io
from util.LocationUtils import locate_byMqtt
class FileUpload(Thread):
__slots__ = ('_fb_queue', '_context', '_image_queue', '_analyse_type', '_msg','_mqtt_list')
__slots__ = ('_fb_queue', '_context', '_image_queue', '_analyse_type', '_msg', '_mqtt_list')
def __init__(self, *args):
super().__init__()
self._fb_queue, self._context, self._msg, self._image_queue, self._analyse_type,self._mqtt_list = args
self._fb_queue, self._context, self._msg, self._image_queue, self._analyse_type, self._mqtt_list = args
self._storage_source = self._context['service']['storage_source']
self._algStatus = False # 默认关闭
self._algStatus = False # 默认关闭
# self._algStatus = True # 默认关闭
self._algSwitch = self._context['service']['algSwitch']
#0521:
default_enabled = str(self._msg.get("defaultEnabled", "True")).lower() == "true"
self._algSwitch = self._context['service']['algSwitch']
# 0521:
default_enabled = str(self._msg.get("defaultEnabled", "True")).lower() == "true"
if default_enabled:
print("执行默认程序（defaultEnabled=True）")
self._algSwitch = True
@ -42,24 +42,16 @@ class FileUpload(Thread):
print("执行替代程序（defaultEnabled=False）")
# 这里放非默认逻辑的代码
self._algSwitch = False
print("---line46 :FileUploadThread.py---",self._algSwitch)
#如果任务是在线、离线处理，则用此类
print("---line46 :FileUploadThread.py---", self._algSwitch)
# 如果任务是在线、离线处理，则用此类
class ImageFileUpload(FileUpload):
__slots__ = ()
#@staticmethod
def handle_image(self,frame_msg, frame_step):
# @staticmethod
def handle_image(self, frame_msg, frame_step):
# (high_score_image["code"], all_frames, draw_config["font_config"])
# high_score_image["code"][code][cls] = (frame, frame_index_list[i], cls_list)
det_xywh, frame, current_frame, all_frames, font_config = frame_msg
@ -72,9 +64,8 @@ class ImageFileUpload(FileUpload):
模型编号：modeCode
检测目标：detectTargetCode
'''
print('*'*100,' mqtt_list:',len(self._mqtt_list))
print('*' * 100, ' mqtt_list:', len(self._mqtt_list))
model_info = []
# 更加模型编码解析数据
for code, det_list in det_xywh.items():
@ -83,15 +74,25 @@ class ImageFileUpload(FileUpload):
if len(target_list) > 0:
aFrame = frame.copy()
for target in target_list:
draw_painting_joint(target[1], aFrame, target[3], target[2], target[4], font_config, target[5])
igH,igW = aFrame.shape[0:2]
if len(self._mqtt_list)>=1:
#camParas = self._mqtt_list[0]['data']
# 自研车牌模型判断
if ModelType.CITY_CARPLATE_MODEL.value[1] == str(code):
draw_name_ocr(target[1], aFrame, target[4])
elif ModelType.CITY_DENSECROWDCOUNT_MODEL.value[1] == str(code) or\
ModelType.CITY_UNDERBUILDCOUNT_MODEL.value[1] == str(code):
draw_name_crowd(target[1], aFrame, target[4])
else:
draw_painting_joint(target[1], aFrame, target[3], target[2], target[4], font_config,
target[5])
igH, igW = aFrame.shape[0:2]
if len(self._mqtt_list) >= 1:
# camParas = self._mqtt_list[0]['data']
camParas = self._mqtt_list[0]
gps = locate_byMqtt(target[1],igW,igH,camParas,outFormat='wgs84')
gps = locate_byMqtt(target[1], igW, igH, camParas, outFormat='wgs84')
else:
gps=[None,None]
model_info.append({"modelCode": str(code), "detectTargetCode": str(cls), "aFrame": aFrame,'gps':gps})
gps = [None, None]
model_info.append(
{"modelCode": str(code), "detectTargetCode": str(cls), "aFrame": aFrame, 'gps': gps})
if len(model_info) > 0:
image_result = {
"or_frame": frame,
@ -110,13 +111,15 @@ class ImageFileUpload(FileUpload):
image_queue, fb_queue, analyse_type = self._image_queue, self._fb_queue, self._analyse_type
service_timeout = int(service["timeout"])
frame_step = int(service["filter"]["frame_step"]) + 120
if msg['taskType']==0: self._algStatus = False
else: self._algStatus = True
if msg['taskType'] == 0:
self._algStatus = False
else:
self._algStatus = True
try:
with ThreadPoolExecutor(max_workers=2) as t:
# 初始化oss客户端
if self._storage_source==1:
minioSdk = MinioSdk(base_dir, env, request_id )
if self._storage_source == 1:
minioSdk = MinioSdk(base_dir, env, request_id)
else:
aliyunOssSdk = AliyunOssSdk(base_dir, env, request_id)
start_time = time()
@ -130,15 +133,17 @@ class ImageFileUpload(FileUpload):
# 获取队列中的消息
image_msg = get_no_block_queue(image_queue)
if image_msg is not None:
if image_msg[0] == 2:
logger.info("图片上传线程收到命令:{}, requestId: {}",image_msg[1] ,request_id)
logger.info("图片上传线程收到命令:{}, requestId: {}", image_msg[1], request_id)
if 'stop' == image_msg[1]:
logger.info("开始停止图片上传线程, requestId:{}", request_id)
break
if 'algStart' == image_msg[1]: self._algStatus = True; logger.info("图片上传线程,执行算法开启命令, requestId:{}", request_id)
if 'algStop' == image_msg[1]: self._algStatus = False; logger.info("图片上传线程,执行算法关闭命令, requestId:{}", request_id)
if 'algStart' == image_msg[1]: self._algStatus = True; logger.info(
"图片上传线程,执行算法开启命令, requestId:{}", request_id)
if 'algStop' == image_msg[1]: self._algStatus = False; logger.info(
"图片上传线程,执行算法关闭命令, requestId:{}", request_id)
if image_msg[0] == 1:
image_result = self.handle_image(image_msg[1], frame_step)
if image_result is not None:
@ -148,8 +153,8 @@ class ImageFileUpload(FileUpload):
image_result["last_frame"],
analyse_type,
"OR", "0", "0", request_id)
if self._storage_source==1:
or_future = t.submit(minioSdk.put_object, or_image,or_image_name)
if self._storage_source == 1:
or_future = t.submit(minioSdk.put_object, or_image, or_image_name)
else:
or_future = t.submit(aliyunOssSdk.put_object, or_image_name, or_image.tobytes())
task.append(or_future)
@ -164,38 +169,38 @@ class ImageFileUpload(FileUpload):
model_info["modelCode"],
model_info["detectTargetCode"],
request_id)
if self._storage_source==1:
if self._storage_source == 1:
ai_future = t.submit(minioSdk.put_object, ai_image,
ai_image_name)
ai_image_name)
else:
ai_future = t.submit(aliyunOssSdk.put_object, ai_image_name,
ai_image.tobytes())
ai_image.tobytes())
task.append(ai_future)
#msg_list.append(message_feedback(request_id,
# msg_list.append(message_feedback(request_id,
# AnalysisStatus.RUNNING.value,
# analyse_type, "", "", "",
# or_image_name,
# ai_image_name,
# model_info['modelCode'],
# model_info['detectTargetCode']))
remote_image_list=[]
remote_image_list = []
for tk in task:
remote_image_list.append( tk.result())
remote_image_list.append(tk.result())
for ii,model_info in enumerate(model_info_list):
msg_list.append( message_feedback(request_id,
for ii, model_info in enumerate(model_info_list):
msg_list.append(message_feedback(request_id,
AnalysisStatus.RUNNING.value,
analyse_type, "", "", "",
remote_image_list[0],
remote_image_list[ii+1],
remote_image_list[ii + 1],
model_info['modelCode'],
model_info['detectTargetCode'],
longitude=model_info['gps'][0],
latitude=model_info['gps'][1],
) )
if (not self._algSwitch) or ( self._algStatus and self._algSwitch):
))
if (not self._algSwitch) or (self._algStatus and self._algSwitch):
for msg in msg_list:
put_queue(fb_queue, msg, timeout=2, is_ex=False)
del task, msg_list
@ -220,9 +225,9 @@ def build_image_name(*args):
time_now = TimeUtils.now_date_to_str("%Y-%m-%d-%H-%M-%S")
return "%s/%s_frame-%s-%s_type_%s-%s-%s-%s_%s.jpg" % (request_id, time_now, current_frame, last_frame,
random_num, mode_type, modeCode, target, image_type)
#如果任务是图像处理，则用此类
# 如果任务是图像处理，则用此类
class ImageTypeImageFileUpload(Thread):
__slots__ = ('_fb_queue', '_context', '_image_queue', '_analyse_type', '_msg')
@ -230,6 +235,7 @@ class ImageTypeImageFileUpload(Thread):
super().__init__()
self._fb_queue, self._context, self._msg, self._image_queue, self._analyse_type = args
self._storage_source = self._context['service']['storage_source']
@staticmethod
def handle_image(det_xywh, copy_frame, font_config):
"""
@ -249,12 +255,21 @@ class ImageTypeImageFileUpload(Thread):
if target_list is not None and len(target_list) > 0:
aiFrame = copy_frame.copy()
for target in target_list:
draw_painting_joint(target[1], aiFrame, target[3], target[2], target[4], font_config)
model_info.append({
"modelCode": str(code),
"detectTargetCode": str(cls),
"frame": aiFrame
})
# 自研车牌模型判断
if ModelType.CITY_CARPLATE_MODEL.value[1] == str(code):
draw_name_ocr(target, aiFrame, font_config[cls])
elif ModelType.CITY_DENSECROWDCOUNT_MODEL.value[1] == str(code) or \
ModelType.CITY_UNDERBUILDCOUNT_MODEL.value[1] == str(code):
draw_name_crowd(target, aiFrame, font_config[cls])
else:
draw_painting_joint(target[1], aiFrame, target[3], target[2], target[4], font_config)
model_info.append({
"modelCode": str(code),
"detectTargetCode": str(cls),
"frame": aiFrame
})
if len(model_info) > 0:
image_result = {
"or_frame": copy_frame,
@ -274,11 +289,11 @@ class ImageTypeImageFileUpload(Thread):
with ThreadPoolExecutor(max_workers=2) as t:
try:
# 初始化oss客户端
if self._storage_source==1:
minioSdk = MinioSdk(base_dir, env, request_id )
if self._storage_source == 1:
minioSdk = MinioSdk(base_dir, env, request_id)
else:
aliyunOssSdk = AliyunOssSdk(base_dir, env, request_id)
start_time = time()
while True:
try:
@ -299,15 +314,15 @@ class ImageTypeImageFileUpload(Thread):
if det_xywh is None:
ai_image_name = build_image_name(0, 0, analyse_type, "AI", result.get("modelCode"),
result.get("type"), request_id)
if self._storage_source==1:
ai_future = t.submit(minioSdk.put_object, copy_frame,ai_image_name)
if self._storage_source == 1:
ai_future = t.submit(minioSdk.put_object, copy_frame, ai_image_name)
else:
ai_future = t.submit(aliyunOssSdk.put_object, ai_image_name, copy_frame)
task.append(ai_future)
remote_names.append(ai_image_name)
#msg_list.append(message_feedback(request_id,
# msg_list.append(message_feedback(request_id,
# AnalysisStatus.RUNNING.value,
# analyse_type, "", "", "",
# image_url,
@ -318,17 +333,17 @@ class ImageTypeImageFileUpload(Thread):
else:
image_result = self.handle_image(det_xywh, copy_frame, font_config)
if image_result:
# 图片帧数编码
if image_url is None:
or_result, or_image = cv2.imencode(".jpg", image_result.get("or_frame"))
image_url_0 = build_image_name(image_result.get("current_frame"),
image_result.get("last_frame"),
analyse_type,
"OR", "0", "O", request_id)
if self._storage_source==1:
or_future = t.submit(minioSdk.put_object, or_image,image_url_0)
image_result.get("last_frame"),
analyse_type,
"OR", "0", "O", request_id)
if self._storage_source == 1:
or_future = t.submit(minioSdk.put_object, or_image, image_url_0)
else:
or_future = t.submit(aliyunOssSdk.put_object, image_url_0,
or_image.tobytes())
@ -344,14 +359,14 @@ class ImageTypeImageFileUpload(Thread):
model_info.get("modelCode"),
model_info.get("detectTargetCode"),
request_id)
if self._storage_source==1:
ai_future = t.submit(minioSdk.put_object, ai_image, ai_image_name)
else:
if self._storage_source == 1:
ai_future = t.submit(minioSdk.put_object, ai_image, ai_image_name)
else:
ai_future = t.submit(aliyunOssSdk.put_object, ai_image_name,
ai_image.tobytes())
task.append(ai_future)
remote_names.append(ai_image_name)
#msg_list.append(message_feedback(request_id,
# msg_list.append(message_feedback(request_id,
# AnalysisStatus.RUNNING.value,
# analyse_type, "", "", "",
# image_url,
@ -362,9 +377,8 @@ class ImageTypeImageFileUpload(Thread):
remote_url_list = []
for thread_result in task:
remote_url_list.append(thread_result.result())
#以下代码是为了获取图像上传后，返回的全路径地址
# 以下代码是为了获取图像上传后，返回的全路径地址
if det_xywh is None:
msg_list.append(message_feedback(request_id,
AnalysisStatus.RUNNING.value,
@ -377,12 +391,12 @@ class ImageTypeImageFileUpload(Thread):
else:
if image_result:
if image_url is None:
for ii in range(len(remote_names)-1):
for ii in range(len(remote_names) - 1):
msg_list.append(message_feedback(request_id,
AnalysisStatus.RUNNING.value,
analyse_type, "", "", "",
remote_url_list[0],
remote_url_list[1+ii],
remote_url_list[1 + ii],
model_info.get('modelCode'),
model_info.get('detectTargetCode'),
analyse_results=result))
@ -394,13 +408,10 @@ class ImageTypeImageFileUpload(Thread):
image_url,
remote_url_list[ii],
model_info_list[ii].get('modelCode'),
model_info_list[ii].get('detectTargetCode'),
model_info_list[ii].get(
'detectTargetCode'),
analyse_results=result))
for msg in msg_list:
put_queue(fb_queue, msg, timeout=2, is_ex=False)
else:

231
concurrency/IntelligentRecognitionProcess.py
View File

@ -62,7 +62,7 @@ class IntelligentRecognitionProcess(Process):
# 发送waitting消息
put_queue(self._fb_queue, message_feedback(self._msg["request_id"], AnalysisStatus.WAITING.value,
self._analyse_type, progress=init_progess), timeout=2, is_ex=True)
self._storage_source = self._context['service']['storage_source']
self._storage_source = self._context['service']['storage_source']
self._algStatus = False
def sendEvent(self, eBody):
put_queue(self.event_queue, eBody, timeout=2, is_ex=True)
@ -91,8 +91,6 @@ class IntelligentRecognitionProcess(Process):
hb_thread.start()
return hb_thread
class OnlineIntelligentRecognitionProcess(IntelligentRecognitionProcess):
@ -113,19 +111,16 @@ class OnlineIntelligentRecognitionProcess(IntelligentRecognitionProcess):
pullProcess.start()
return pullProcess
def upload_video(self,base_dir, env, request_id, orFilePath, aiFilePath):
if self._storage_source==1:
minioSdk = MinioSdk(base_dir, env, request_id )
minioSdk = MinioSdk(base_dir, env, request_id)
upload_video_thread_or = Common(minioSdk.put_object, orFilePath, "or_online_%s.mp4" % request_id)
upload_video_thread_ai = Common(minioSdk.put_object, aiFilePath, "ai_online_%s.mp4" % request_id)
else:
else:
aliyunVodSdk = ThAliyunVodSdk(base_dir, env, request_id)
upload_video_thread_or = Common(aliyunVodSdk.get_play_url, orFilePath, "or_online_%s" % request_id)
upload_video_thread_ai = Common(aliyunVodSdk.get_play_url, aiFilePath, "ai_online_%s" % request_id)
upload_video_thread_or.setDaemon(True)
upload_video_thread_ai.setDaemon(True)
upload_video_thread_or.start()
@ -146,7 +141,7 @@ class OnlineIntelligentRecognitionProcess(IntelligentRecognitionProcess):
or_url = upload_video_thread_or.get_result()
ai_url = upload_video_thread_ai.get_result()
return or_url, ai_url
'''
'''
@staticmethod
def ai_normal_dtection(model, frame, request_id):
@ -226,10 +221,10 @@ class OnlineIntelligentRecognitionProcess(IntelligentRecognitionProcess):
ex = None
# 拉流进程、推流进程、心跳线程
pull_process, push_process, hb_thread = None, None, None
# 事件队列、拉流队列、心跳队列、反馈队列
event_queue, pull_queue, hb_queue, fb_queue = self.event_queue, self._pull_queue, self._hb_queue, self._fb_queue
# 推流队列、推流异常队列、图片队列
push_queue, push_ex_queue, image_queue = self._push_queue, self._push_ex_queue, self._image_queue
try:
@ -237,19 +232,18 @@ class OnlineIntelligentRecognitionProcess(IntelligentRecognitionProcess):
init_log(base_dir, env)
# 打印启动日志
logger.info("开始启动实时分析进程！requestId: {}", request_id)
# 启动拉流进程(包含拉流线程, 图片上传线程,mqtt读取线程)
# 拉流进程初始化时间长, 先启动
pull_process = self.start_pull_stream(msg, context, fb_queue, pull_queue, image_queue, analyse_type, 25)
#print_cpu_status(requestId=request_id,lineNum=inspect.currentframe().f_lineno) #7.0,
#print_cpu_status(requestId=request_id,lineNum=inspect.currentframe().f_lineno) #7.0,
# 启动心跳线程
hb_thread = self.start_heartbeat(fb_queue, hb_queue, request_id, analyse_type, context)
#print_cpu_status(requestId=request_id,lineNum=inspect.currentframe().f_lineno) #7.0,
# print_cpu_status(requestId=request_id,lineNum=inspect.currentframe().f_lineno) #7.0,
# 加载算法模型
model_array = get_model(msg, context, analyse_type)
#print_cpu_status(requestId=request_id,lineNum=inspect.currentframe().f_lineno) #9.5
# print_cpu_status(requestId=request_id,lineNum=inspect.currentframe().f_lineno) #9.5
# 启动推流进程
push_process = self.start_push_stream(msg, push_queue, image_queue, push_ex_queue, hb_queue, context)
#print_cpu_status(requestId=request_id,lineNum=inspect.currentframe().f_lineno)
@ -273,7 +267,7 @@ class OnlineIntelligentRecognitionProcess(IntelligentRecognitionProcess):
raise ServiceException(push_status[1], push_status[2])
# 获取停止指令
event_result = get_no_block_queue(event_queue)
if event_result:
cmdStr = event_result.get("command")
#接收到算法开启、或者关闭的命令
@ -281,7 +275,7 @@ class OnlineIntelligentRecognitionProcess(IntelligentRecognitionProcess):
logger.info("发送向推流进程发送算法命令, requestId: {}, {}", request_id,cmdStr )
put_queue(push_queue, (2, cmdStr), timeout=1, is_ex=True)
pull_process.sendCommand({"command": cmdStr})
# 接收到停止指令
if "stop" == cmdStr:
logger.info("实时任务开始停止, requestId: {}", request_id)
@ -301,20 +295,32 @@ class OnlineIntelligentRecognitionProcess(IntelligentRecognitionProcess):
task_status[0] = 1
for i, model in enumerate(model_array):
model_conf, code = model
model_param = model_conf[1]
# (modeType, model_param, allowedList, names, rainbows)
MODEL_CONFIG[code][2](frame_list[0].shape[1], frame_list[0].shape[0],
model_conf)
if draw_config.get("font_config") is None:
draw_config["font_config"] = model_param['font_config']
if draw_config.get(code) is None:
draw_config[code] = {}
draw_config[code]["allowedList"] = model_conf[2]
draw_config[code]["rainbows"] = model_conf[4]
draw_config[code]["label_arrays"] = model_param['label_arraylist']
if "label_dict" in model_param:
draw_config[code]["label_dict"] = model_param['label_dict']
#print_cpu_status(requestId=request_id,lineNum=inspect.currentframe().f_lineno)
if ModelType.CITY_CARPLATE_MODEL.value[1] == str(code) or \
ModelType.CITY_DENSECROWDCOUNT_MODEL.value[1] == str(code) or\
ModelType.CITY_UNDERBUILDCOUNT_MODEL.value[1] == str(code):
if draw_config.get(code) is None:
draw_config[code] = {}
draw_config["font_config"] = model_conf[4]
draw_config[code]["allowedList"] = 0
draw_config[code]["label_arrays"] = [None]
draw_config[code]["rainbows"] = model_conf[4]
else:
model_param = model_conf[1]
# (modeType, model_param, allowedList, names, rainbows)
MODEL_CONFIG[code][2](frame_list[0].shape[1], frame_list[0].shape[0],
model_conf)
if draw_config.get("font_config") is None:
draw_config["font_config"] = model_param['font_config']
if draw_config.get(code) is None:
draw_config[code] = {}
draw_config[code]["allowedList"] = model_conf[2]
draw_config[code]["rainbows"] = model_conf[4]
draw_config[code]["label_arrays"] = model_param['label_arraylist']
if "label_dict" in model_param:
draw_config[code]["label_dict"] = model_param['label_dict']
# print_cpu_status(requestId=request_id,lineNum=inspect.currentframe().f_lineno)
# 多线程并发处理, 经过测试两个线程最优
det_array = []
for i, frame in enumerate(frame_list):
@ -322,11 +328,11 @@ class OnlineIntelligentRecognitionProcess(IntelligentRecognitionProcess):
frame_index_list[i], tt, request_id)
det_array.append(det_result)
push_objs = [det.result() for det in det_array]
#print_cpu_status(requestId=request_id,lineNum=inspect.currentframe().f_lineno)
# print_cpu_status(requestId=request_id,lineNum=inspect.currentframe().f_lineno)
put_queue(push_queue,
(1, (frame_list, frame_index_list, all_frames, draw_config, push_objs)),
timeout=2, is_ex=True)
#print_cpu_status(requestId=request_id,lineNum=inspect.currentframe().f_lineno)
# print_cpu_status(requestId=request_id,lineNum=inspect.currentframe().f_lineno)
del det_array, push_objs
del frame_list, frame_index_list, all_frames
elif pull_result[0] == 1:
@ -437,23 +443,23 @@ class OnlineIntelligentRecognitionProcess(IntelligentRecognitionProcess):
class OfflineIntelligentRecognitionProcess(IntelligentRecognitionProcess):
__slots__ = ()
def upload_video(self,base_dir, env, request_id, aiFilePath):
aliyunVodSdk = ThAliyunVodSdk(base_dir, env, request_id)
upload_video_thread_ai = Common(aliyunVodSdk.get_play_url, aiFilePath, "ai_online_%s" % request_id)
if self._storage_source==1:
minioSdk = MinioSdk(base_dir, env, request_id )
upload_video_thread_ai = Common(minioSdk.put_object, aiFilePath, "ai_online_%s.mp4" % request_id)
else:
aliyunVodSdk = ThAliyunVodSdk(base_dir, env, request_id)
upload_video_thread_ai = Common(aliyunVodSdk.get_play_url, aiFilePath, "ai_online_%s" % request_id)
upload_video_thread_ai.setDaemon(True)
upload_video_thread_ai.start()
ai_url = upload_video_thread_ai.get_result()
return ai_url
'''
@staticmethod
def upload_video(base_dir, env, request_id, aiFilePath):
@ -602,7 +608,7 @@ class OfflineIntelligentRecognitionProcess(IntelligentRecognitionProcess):
logger.info("发送向推流进程发送算法命令, requestId: {}, {}", request_id,cmdStr )
put_queue(push_queue, (2, cmdStr), timeout=1, is_ex=True)
pull_process.sendCommand({"command": cmdStr})
pull_result = get_no_block_queue(pull_queue)
if pull_result is None:
sleep(1)
@ -616,19 +622,32 @@ class OfflineIntelligentRecognitionProcess(IntelligentRecognitionProcess):
task_status[0] = 1
for i, model in enumerate(model_array):
model_conf, code = model
model_param = model_conf[1]
# (modeType, model_param, allowedList, names, rainbows)
MODEL_CONFIG[code][2](frame_list[0].shape[1], frame_list[0].shape[0],
model_conf)
if draw_config.get("font_config") is None:
draw_config["font_config"] = model_param['font_config']
if draw_config.get(code) is None:
draw_config[code] = {}
draw_config[code]["allowedList"] = model_conf[2]
draw_config[code]["rainbows"] = model_conf[4]
draw_config[code]["label_arrays"] = model_param['label_arraylist']
if "label_dict" in model_param:
draw_config[code]["label_dict"] = model_param['label_dict']
if ModelType.CITY_CARPLATE_MODEL.value[1] == str(code) or \
ModelType.CITY_DENSECROWDCOUNT_MODEL.value[1] == str(code) or\
ModelType.CITY_UNDERBUILDCOUNT_MODEL.value[1] == str(code):
if draw_config.get(code) is None:
draw_config[code] = {}
draw_config["font_config"] = model_conf[4]
draw_config[code]["allowedList"] = 0
draw_config[code]["label_arrays"] = [None]
draw_config[code]["rainbows"] = model_conf[4]
else:
model_param = model_conf[1]
# (modeType, model_param, allowedList, names, rainbows)
MODEL_CONFIG[code][2](frame_list[0].shape[1], frame_list[0].shape[0],
model_conf)
if draw_config.get("font_config") is None:
draw_config["font_config"] = model_param['font_config']
if draw_config.get(code) is None:
draw_config[code] = {}
draw_config[code]["allowedList"] = model_conf[2]
draw_config[code]["rainbows"] = model_conf[4]
draw_config[code]["label_arrays"] = model_param['label_arraylist']
if "label_dict" in model_param:
draw_config[code]["label_dict"] = model_param['label_dict']
det_array = []
for i, frame in enumerate(frame_list):
det_result = t.submit(self.obj_det, self, model_array, frame, task_status,
@ -745,7 +764,7 @@ class PhotosIntelligentRecognitionProcess(Process):
put_queue(self._fb_queue, message_feedback(self._msg["request_id"], AnalysisStatus.WAITING.value,
self._analyse_type, progress=init_progess), timeout=2, is_ex=True)
self.build_logo(self._msg, self._context)
self._storage_source = self._context['service']['storage_source']
self._storage_source = self._context['service']['storage_source']
@staticmethod
def build_logo(msg, context):
@ -922,6 +941,62 @@ class PhotosIntelligentRecognitionProcess(Process):
logger.error("模型分析异常: {}, requestId: {}", format_exc(), request_id)
raise e
#自研车牌模型
def carplate_rec(self, imageUrl, mod, image_queue, request_id):
try:
# model_conf： modeType, allowedList, detpar, ocrmodel, rainbows
model_conf, code = mod
modeType, device, modelList, detpar, rainbows = model_conf
image = url2Array(imageUrl)
dets = {code: {}}
# param = [image, new_device, model, par, img_type, request_id]
# model_conf, frame, device, requestId
dataBack = MODEL_CONFIG[code][3]([[modeType, device, modelList, detpar], image, request_id])[0][2]
dets[code][0] = dataBack
if not dataBack:
logger.info("车牌识别为空")
# for ai_result in dataBack:
# label, box = ai_result
# color = rainbows
if len(dataBack) > 0:
put_queue(image_queue, (1, (dets, imageUrl, image, rainbows, "")), timeout=2, is_ex=False)
except ServiceException as s:
raise s
except Exception as e:
logger.error("模型分析异常: {}, requestId: {}", format_exc(), request_id)
raise e
#密集人群计数
def denscrowdcount_rec(self, imageUrl, mod, image_queue, request_id):
try:
# model_conf： modeType, allowedList, detpar, ocrmodel, rainbows
model_conf, code = mod
modeType, device, model, postPar, rainbows = model_conf
image = url2Array(imageUrl)
dets = {code: {}}
# param = [image, new_device, model, par, img_type, request_id]
# model_conf, frame, device, requestId
dataBack = MODEL_CONFIG[code][3]([[modeType, device, model, postPar], image, request_id])[0][2]
dets[code][0] = dataBack
if not dataBack:
logger.info("当前页面无人")
# for ai_result in dataBack:
# label, box = ai_result
# color = rainbows
if len(dataBack) > 0:
put_queue(image_queue, (1, (dets, imageUrl, image, rainbows, '')), timeout=2, is_ex=False)
except ServiceException as s:
raise s
except Exception as e:
logger.error("模型分析异常: {}, requestId: {}", format_exc(), request_id)
raise e
'''
# 防疫模型
'''
@ -936,6 +1011,26 @@ class PhotosIntelligentRecognitionProcess(Process):
for r in obj_list:
r.result(60)
# 自研车牌识别模型：
def carpalteRec(self, imageUrls, model, image_queue, request_id):
with ThreadPoolExecutor(max_workers=2) as t:
obj_list = []
for imageUrl in imageUrls:
obj = t.submit(self.carplate_rec, imageUrl, model, image_queue, request_id)
obj_list.append(obj)
for r in obj_list:
r.result(60)
# 密集人群计数：CITY_DENSECROWDCOUNT_MODEL
def denscrowdcountRec(self, imageUrls, model, image_queue, request_id):
with ThreadPoolExecutor(max_workers=2) as t:
obj_list = []
for imageUrl in imageUrls:
obj = t.submit(self.denscrowdcount_rec, imageUrl, model, image_queue, request_id)
obj_list.append(obj)
for r in obj_list:
r.result(60)
def image_recognition(self, imageUrl, mod, image_queue, logo, request_id):
try:
model_conf, code = mod
@ -1125,7 +1220,7 @@ class PhotosIntelligentRecognitionProcess(Process):
except requests.exceptions.RequestException as e:
# 捕获请求过程中可能出现的异常（如网络问题、超时等）
return False,str(e)
def run(self):
fb_queue, msg, analyse_type, context = self._fb_queue, self._msg, self._analyse_type, self._context
request_id, logo, image_queue = msg["request_id"], context['logo'], self._image_queue
@ -1136,7 +1231,7 @@ class PhotosIntelligentRecognitionProcess(Process):
valFlag=True
for url in imageUrls:
valFlag,ret = self.check_ImageUrl_Vaild(url,timeout=1)
if not valFlag:
logger.error("图片分析异常: {}, requestId:{},url:{}",ret, request_id,url)
#print("AnalysisStatus.FAILED.value:{}，ExceptionType.URL_ADDRESS_ACCESS_FAILED.value[0]:{},ExceptionType.URL_ADDRESS_ACCESS_FAILED.value[1]:{}".format(AnalysisStatus.FAILED.value,ExceptionType.URL_ADDRESS_ACCESS_FAILED.value[0],ExceptionType.URL_ADDRESS_ACCESS_FAILED.value[1] ) )
@ -1168,6 +1263,15 @@ class PhotosIntelligentRecognitionProcess(Process):
elif model[1] == ModelType.PLATE_MODEL.value[1]:
result = t.submit(self.epidemicPrevention, imageUrls, model, base_dir, env, request_id)
task_list.append(result)
# 自研车牌模型
elif model[1] == ModelType.CITY_CARPLATE_MODEL.value[1]:
result = t.submit(self.carpalteRec, imageUrls, model, image_queue, request_id)
task_list.append(result)
# 人群计数模型
elif model[1] == ModelType.CITY_DENSECROWDCOUNT_MODEL.value[1] or \
model[1] == ModelType.CITY_UNDERBUILDCOUNT_MODEL.value[1]:
result = t.submit(self.denscrowdcountRec, imageUrls, model, image_queue, request_id)
task_list.append(result)
else:
result = t.submit(self.publicIdentification, imageUrls, model, image_queue, logo, request_id)
task_list.append(result)
@ -1214,7 +1318,7 @@ class ScreenRecordingProcess(Process):
put_queue(self._fb_queue,
recording_feedback(self._msg["request_id"], RecordingStatus.RECORDING_WAITING.value[0]),
timeout=1, is_ex=True)
self._storage_source = self._context['service']['storage_source']
self._storage_source = self._context['service']['storage_source']
def sendEvent(self, result):
put_queue(self._event_queue, result, timeout=2, is_ex=True)
@ -1380,17 +1484,14 @@ class ScreenRecordingProcess(Process):
clear_queue(self._hb_queue)
clear_queue(self._pull_queue)
def upload_video(self,base_dir, env, request_id, orFilePath):
if self._storage_source==1:
minioSdk = MinioSdk(base_dir, env, request_id )
def upload_video(self, base_dir, env, request_id, orFilePath):
if self._storage_source == 1:
minioSdk = MinioSdk(base_dir, env, request_id)
upload_video_thread_ai = Common(minioSdk.put_object, aiFilePath, "%s/ai_online.mp4" % request_id)
else:
aliyunVodSdk = ThAliyunVodSdk(base_dir, env, request_id)
upload_video_thread_ai = Common(aliyunVodSdk.get_play_url, aiFilePath, "ai_online_%s" % request_id)
upload_video_thread_ai.setDaemon(True)
upload_video_thread_ai.start()
or_url = upload_video_thread_ai.get_result()

136
concurrency/PushVideoStreamProcess.py
View File

@ -23,7 +23,7 @@ from util.Cv2Utils import video_conjuncing, write_or_video, write_ai_video, push
from util.ImageUtils import url2Array, add_water_pic
from util.LogUtils import init_log
from util.PlotsUtils import draw_painting_joint, filterBox, xywh2xyxy2, draw_name_joint
from util.PlotsUtils import draw_painting_joint, filterBox, xywh2xyxy2, xy2xyxy, draw_name_joint, plot_one_box_auto, draw_name_ocr,draw_name_crowd
from util.QueUtil import get_no_block_queue, put_queue, clear_queue
@ -37,7 +37,7 @@ class PushStreamProcess(Process):
self._msg, self._push_queue, self._image_queue, self._push_ex_queue, self._hb_queue, self._context = args
self._algStatus = False # 默认关闭
self._algSwitch = self._context['service']['algSwitch']
#0521:
default_enabled = str(self._msg.get("defaultEnabled", "True")).lower() == "true"
@ -49,16 +49,9 @@ class PushStreamProcess(Process):
print("执行替代程序（defaultEnabled=False）")
# 这里放非默认逻辑的代码
self._algSwitch = False
print("---line53 :PushVideoStreamProcess.py---",self._algSwitch)
def build_logo_url(self):
logo = None
if self._context["video"]["video_add_water"]:
@ -155,17 +148,38 @@ class OnPushStreamProcess(PushStreamProcess):
font_config, allowedList = draw_config["font_config"], draw_config[code]["allowedList"]
rainbows, label_arrays = draw_config[code]["rainbows"], draw_config[code]["label_arrays"]
for qs in det_result:
try: # 应对NaN情况
box, score, cls = xywh2xyxy2(qs)
except:
continue
if cls not in allowedList or score < frame_score:
continue
label_array, color = label_arrays[cls], rainbows[cls]
if ModelType.CHANNEL2_MODEL.value[1] == str(code) and cls == 2:
rr = t.submit(draw_name_joint, box, copy_frame, draw_config[code]["label_dict"], score, color, font_config, qs[6])
# 自研车牌模型处理
if ModelType.CITY_CARPLATE_MODEL.value[1] == str(code):
cls = 0
box = xy2xyxy(qs[1])
score = None
color = rainbows[cls]
label_array = None
rr = t.submit(draw_name_ocr, qs, copy_frame, color)
elif ModelType.CITY_DENSECROWDCOUNT_MODEL.value[1] == str(code) or\
ModelType.CITY_UNDERBUILDCOUNT_MODEL.value[1] == str(code):
cls = 0
box = [(0, 0), (0, 0), (0, 0), (0, 0)]
score = None
color = rainbows[cls]
label_array = None
rr = t.submit(draw_name_crowd, qs, copy_frame, color)
else:
rr = t.submit(draw_painting_joint, box, copy_frame, label_array, score, color, font_config)
try: # 应对NaN情况
box, score, cls = xywh2xyxy2(qs)
except:
continue
if cls not in allowedList or score < frame_score:
continue
label_array, color = label_arrays[cls], rainbows[cls]
if ModelType.CHANNEL2_MODEL.value[1] == str(code) and cls == 2:
rr = t.submit(draw_name_joint, box, copy_frame,
draw_config[code]["label_dict"], score, color,
font_config, qs[6])
else:
rr = t.submit(draw_painting_joint, box, copy_frame, label_array,
score, color, font_config)
thread_p.append(rr)
if det_xywh.get(code) is None:
det_xywh[code] = {}
@ -184,17 +198,17 @@ class OnPushStreamProcess(PushStreamProcess):
box[2][0], box[2][1], box[3][0], box[3][1],
score, cls, code],dtype=np.float32)
qs_np = np.row_stack((qs_np, result_li))
if logo:
frame = add_water_pic(frame, logo, request_id)
copy_frame = add_water_pic(copy_frame, logo, request_id)
if len(thread_p) > 0:
for r in thread_p:
r.result()
#print('----line173:',self._algSwitch,self._algStatus)
#print('----line173:',self._algSwitch,self._algStatus)
if self._algSwitch and (not self._algStatus):
frame_merge = video_conjuncing(frame, frame.copy())
else:
else:
frame_merge = video_conjuncing(frame, copy_frame)
# 写原视频到本地
write_or_video_result = t.submit(write_or_video, frame, orFilePath, or_video_file,
@ -207,7 +221,7 @@ class OnPushStreamProcess(PushStreamProcess):
# 如果有问题, 走下面的逻辑
if qs_np is not None:
if len(qs_np.shape) == 1:
qs_np = qs_np[np.newaxis,...]
qs_np = qs_np[np.newaxis,...]
qs_np_id = qs_np.copy()
b = np.ones(qs_np_id.shape[0])
qs_np_id = np.column_stack((qs_np_id,b))
@ -233,7 +247,7 @@ class OnPushStreamProcess(PushStreamProcess):
if q[11] >= 1:
cls = int(q[9])
if not (cls in new_lab):
continue # 为了防止其他类别被带出
continue # 为了防止其他类别被带出
code = str(int(q[10])).zfill(3)
if det_xywh2.get(code) is None:
det_xywh2[code] = {}
@ -246,8 +260,12 @@ class OnPushStreamProcess(PushStreamProcess):
is_new = False
if q[11] == 1:
is_new = True
if ModelType.CITY_CARPLATE_MODEL.value[1] == str(code) or \
ModelType.CITY_DENSECROWDCOUNT_MODEL.value[1] == str(code) or\
ModelType.CITY_UNDERBUILDCOUNT_MODEL.value[1] == str(code):
box = qs
if cd is None:
det_xywh2[code][cls] = [[cls, box, score, label_array, color, is_new]]
det_xywh2[code][cls] = [[cls, box, score, label_array, color, is_new]]
else:
det_xywh2[code][cls].append([cls, box, score, label_array, color, is_new])
if len(det_xywh2) > 0:
@ -268,7 +286,7 @@ class OnPushStreamProcess(PushStreamProcess):
ex_status = False
logger.info("停止推流进程, requestId: {}", request_id)
break
del push_r
else:
sleep(1)
@ -363,24 +381,44 @@ class OffPushStreamProcess(PushStreamProcess):
qs_reurn = []
for det in push_objs[i]:
code, det_result = det
# 每个单独模型处理
# 模型编号、100帧的所有问题, 检测目标、颜色、文字图片
if len(det_result) > 0:
font_config, allowedList = draw_config["font_config"], draw_config[code]["allowedList"]
rainbows, label_arrays = draw_config[code]["rainbows"], draw_config[code]["label_arrays"]
for qs in det_result:
box, score, cls = xywh2xyxy2(qs)
if cls not in allowedList or score < frame_score:
continue
label_array, color = label_arrays[cls], rainbows[cls]
if ModelType.CHANNEL2_MODEL.value[1] == str(code) and cls == 2:
rr = t.submit(draw_name_joint, box, copy_frame, draw_config[code]["label_dict"], score, color, font_config, qs[6])
# 自研车牌模型处理
if ModelType.CITY_CARPLATE_MODEL.value[1] == str(code):
cls = 0
box = xy2xyxy(qs[1])
score = None
color = rainbows[cls]
label_array = None
label_arrays = [None]
rr = t.submit(draw_name_ocr, qs, copy_frame, color)
elif ModelType.CITY_DENSECROWDCOUNT_MODEL.value[1] == str(code) or\
ModelType.CITY_UNDERBUILDCOUNT_MODEL.value[1] == str(code):
cls = 0
box = [(0,0),(0,0),(0,0),(0,0)]
score = None
color = rainbows[cls]
label_array = None
rr = t.submit(draw_name_crowd, qs, copy_frame, color)
else:
rr = t.submit(draw_painting_joint, box, copy_frame, label_array, score, color, font_config)
box, score, cls = xywh2xyxy2(qs)
if cls not in allowedList or score < frame_score:
continue
label_array, color = label_arrays[cls], rainbows[cls]
if ModelType.CHANNEL2_MODEL.value[1] == str(code) and cls == 2:
rr = t.submit(draw_name_joint, box, copy_frame, draw_config[code]["label_dict"], score, color, font_config, qs[6])
else:
rr = t.submit(draw_painting_joint, box, copy_frame, label_array, score, color, font_config)
thread_p.append(rr)
if det_xywh.get(code) is None:
det_xywh[code] = {}
cd = det_xywh[code].get(cls)
@ -388,17 +426,17 @@ class OffPushStreamProcess(PushStreamProcess):
if cd is None:
det_xywh[code][cls] = [[cls, box, score, label_array, color]]
else:
det_xywh[code][cls].append([cls, box, score, label_array, color])
det_xywh[code][cls].append([cls, box, score, label_array, color])
if qs_np is None:
qs_np = np.array([box[0][0], box[0][1], box[1][0], box[1][1],
box[2][0], box[2][1], box[3][0], box[3][1],
qs_np = np.array([box[0][0], box[0][1], box[1][0], box[1][1],
box[2][0], box[2][1], box[3][0], box[3][1],
score, cls, code],dtype=np.float32)
else:
result_li = np.array([box[0][0], box[0][1], box[1][0], box[1][1],
result_li = np.array([box[0][0], box[0][1], box[1][0], box[1][1],
box[2][0], box[2][1], box[3][0], box[3][1],
score, cls, code],dtype=np.float32)
qs_np = np.row_stack((qs_np, result_li))
if logo:
frame = add_water_pic(frame, logo, request_id)
copy_frame = add_water_pic(copy_frame, logo, request_id)
@ -407,7 +445,7 @@ class OffPushStreamProcess(PushStreamProcess):
r.result()
if self._algSwitch and (not self._algStatus):
frame_merge = video_conjuncing(frame, frame.copy())
else:
else:
frame_merge = video_conjuncing(frame, copy_frame)
# 写识别视频到本地
write_ai_video_result = t.submit(write_ai_video, frame_merge, aiFilePath,
@ -416,10 +454,9 @@ class OffPushStreamProcess(PushStreamProcess):
push_stream_result = t.submit(push_video_stream, frame_merge, push_p, push_url,
p_push_status, request_id)
if qs_np is not None:
if len(qs_np.shape) == 1:
qs_np = qs_np[np.newaxis,...]
qs_np = qs_np[np.newaxis,...]
qs_np_id = qs_np.copy()
b = np.ones(qs_np_id.shape[0])
qs_np_id = np.column_stack((qs_np_id,b))
@ -446,7 +483,7 @@ class OffPushStreamProcess(PushStreamProcess):
if q[11] >= 1:
cls = int(q[9])
if not (cls in new_lab):
continue # 为了防止其他类别被带出
continue # 为了防止其他类别被带出
code = str(int(q[10])).zfill(3)
if det_xywh2.get(code) is None:
det_xywh2[code] = {}
@ -459,6 +496,11 @@ class OffPushStreamProcess(PushStreamProcess):
is_new = False
if q[11] == 1:
is_new = True
if ModelType.CITY_CARPLATE_MODEL.value[1] == str(code) or \
ModelType.CITY_DENSECROWDCOUNT_MODEL.value[1] == str(code) or\
ModelType.CITY_UNDERBUILDCOUNT_MODEL.value[1] == str(code):
box = qs
if cd is None:
det_xywh2[code][cls] = [[cls, box, score, label_array, color, is_new]]
else:

267
enums/ModelTypeEnum.py
View File

@ -9,15 +9,14 @@ from DMPRUtils.jointUtil import dmpr_yolo
from segutils.segmodel import SegModel
from utilsK.queRiver import riverDetSegMixProcess
from utilsK.crowdGather import gather_post_process
from segutils.trafficUtils import tracfficAccidentMixFunction
from segutils.trafficUtils import tracfficAccidentMixFunction,mixTraffic_postprocess
from utilsK.drownUtils import mixDrowing_water_postprocess
from utilsK.noParkingUtils import mixNoParking_road_postprocess
from utilsK.illParkingUtils import illParking_postprocess
from utilsK.spillUtils import mixSpillage_postprocess
from utilsK.cthcUtils import mixCthc_postprocess
from utilsK.pannelpostUtils import pannel_post_process
from stdc import stdcModel
from yolov5 import yolov5Model
from p2pNet import p2NnetModel
from DMPRUtils.jointUtil import dmpr_yolo_stdc
from AI import default_mix
from ocr import ocrModel
@ -68,7 +67,7 @@ class ModelType(Enum):
'Detweights': "../weights/trt/AIlib2/river/yolov5_%s_fp16.engine" % gpuName,
'Segweights': '../weights/trt/AIlib2/river/stdc_360X640_%s_fp16.engine' % gpuName
})
# FOREST_FARM_MODEL = ("2", "002", "森林模型", 'forest2', lambda device, gpuName: {
# 'device': device,
# 'gpu_name': gpuName,
@ -89,7 +88,7 @@ class ModelType(Enum):
# },
# 'Segweights': None
# })
FOREST_FARM_MODEL = ("2", "002", "森林模型", 'forest2', lambda device, gpuName: {
'labelnames': ["林斑", "病死树", "行人", "火焰", "烟雾","云朵"],
@ -102,10 +101,10 @@ class ModelType(Enum):
'model':yolov5Model,
'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 } },
}
],
'postFile': {
"name": "post_process",
"conf_thres": 0.25,
@ -117,9 +116,8 @@ class ModelType(Enum):
'segRegionCnt':2,###分割模型结果需要保留的等值线数目
"pixScale": 1.2,
})
TRAFFIC_FARM_MODEL = ("3", "003", "交通模型", 'highWay2', lambda device, gpuName: {
@ -131,8 +129,7 @@ class ModelType(Enum):
'seg_nclass': 3,
'segRegionCnt': 2,
'segPar': {
#'modelSize': (640, 360),
'modelSize': (1920, 1080),
'modelSize': (640, 360),
'mean': (0.485, 0.456, 0.406),
'std': (0.229, 0.224, 0.225),
'predResize': True,
@ -141,8 +138,7 @@ class ModelType(Enum):
'mixFunction': {
'function': tracfficAccidentMixFunction,
'pars': {
#'modelSize': (640, 360),
'modelSize': (1920,1080),
'modelSize': (640, 360),
'RoadArea': 16000,
'roadVehicleAngle': 15,
'speedRoadVehicleAngleMax': 75,
@ -166,7 +162,7 @@ class ModelType(Enum):
"classes": 10,
"rainbows": COLOR
},
'allowedList':[0,1,2,3,4,5,6,7,8,9,10,11,12,16,17,18,19,20,21,22],
'allowedList':[0,1,2,3,4,5,6,7,8,9,10,11,12,16,17,18,19,20,21,22],
'Detweights': "../weights/trt/AIlib2/highWay2/yolov5_%s_fp16.engine" % gpuName,
'Segweights': '../weights/trt/AIlib2/highWay2/stdc_360X640_%s_fp16.engine' % gpuName
})
@ -361,30 +357,27 @@ class ModelType(Enum):
"classes": 5,
"rainbows": COLOR
},
# "../weights/pth/AIlib2/%s/yolov5.pt" % modeType.value[3]
'Detweights': "../weights/trt/AIlib2/river2/yolov5_%s_fp16.engine" % gpuName,
# '../weights/pth/AIlib2/%s/stdc_360X640.pth' % modeType.value[3]
'Segweights': '../weights/trt/AIlib2/river2/stdc_360X640_%s_fp16.engine' % gpuName
})
CITY_MANGEMENT_MODEL = ("16", "016", "城管模型", 'cityMangement2', lambda device, gpuName: {
'labelnames': [ "车辆", "垃圾", "商贩", "违停","占道经营","裸土","未覆盖裸土" ],
'labelnames': [ "车辆", "垃圾", "商贩", "违停","占道经营","裸土","未覆盖裸土","违建" ],
'postProcess':{
'function':dmpr_yolo_stdc,
'pars':{
'carCls':0 ,'illCls':6,'scaleRatio':0.5,'border':80,
#车辆","垃圾","商贩","裸土","占道经营","违停"--->
#"车辆","垃圾","商贩","违停","占道经营","裸土"
'classReindex':{ 0:0,1:1,2:2,3:6,4:4,5:5,6:3}
'carCls':0 ,'illCls':7,'scaleRatio':0.5,'border':80,
#"车辆","垃圾","商贩","裸土","占道经营","未覆盖裸土","违建"
# key:实际训练index value:展示index
'classReindex':{ 0:0,1:1,2:2,7:3,4:4,3:5,5:6,6:7}
}
},
'models':[
{
'weight':'../weights/pth/AIlib2/cityMangement3/yolov5.pt',
#'weight':'../weights/trt/AIlib2/cityMangement3/yolov5_%s_fp16.engine'%(gpuName),
'name':'yolov5',
'weight':'../weights/trt/AIlib2/cityMangement3/yolov5_%s_fp16.engine'%(gpuName),
'name':'yolov5',
'model':yolov5Model,
'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":0.4,"2":0.5,"3":0.5 } }
'par':{ 'half':True,'device':'cuda:0' ,'conf_thres':0.25,'iou_thres':0.45,'allowedList':[0,1,2,3,4,5,6,7],'segRegionCnt':1, 'trtFlag_det':True,'trtFlag_seg':True, "score_byClass":{"0":0.8,"1":0.4,"2":0.5,"3":0.5 } }
},
{
'weight':'../weights/pth/AIlib2/cityMangement3/dmpr.pth',
@ -392,25 +385,25 @@ class ModelType(Enum):
'depth_factor':32,'NUM_FEATURE_MAP_CHANNEL':6,'dmpr_thresh':0.1, 'dmprimg_size':640,
'name':'dmpr'
},
'model':DMPRModel,
'model':DMPRModel,
'name':'dmpr'
},
{
'weight':'../weights/pth/AIlib2/cityMangement3/stdc_360X640.pth',
{
'weight':'../weights/trt/AIlib2/cityMangement3/stdc_360X640_%s_fp16.engine'%(gpuName),
'par':{
'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
'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
'model':stdcModel,
'name':'stdc'
}
'name':'stdc'
}
],
'postFile': {
"name": "post_process",
"conf_thres": 0.25,
"iou_thres": 0.45,
"classes": 6,
"classes": 8,
"rainbows": COLOR
},
'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
'detModelpara':[{"id":str(x),"config":{"k1":"v1","k2":"v2"}} for x in [0,1,2,3,4,5,6,7,8,9] ],###控制哪些检测类别显示、输出
'segRegionCnt':2,###分割模型结果需要保留的等值线数目
"pixScale": 1.2,
})
@ -443,9 +436,7 @@ class ModelType(Enum):
"classes": 9,
"rainbows": COLOR
},
# "../weights/pth/AIlib2/%s/yolov5.pt" % modeType.value[3]
'Detweights': "../weights/trt/AIlib2/drowning/yolov5_%s_fp16.engine" % gpuName,
# '../weights/pth/AIlib2/%s/stdc_360X640.pth' % modeType.value[3]
'Segweights': '../weights/trt/AIlib2/drowning/stdc_360X640_%s_fp16.engine' % gpuName
})
@ -514,7 +505,7 @@ class ModelType(Enum):
CITYROAD_MODEL = ("20", "020", "城市公路模型", 'cityRoad', lambda device, gpuName: {
'device': device,
'labelnames': ["护栏", "交通标志", "非交通标志", "施工", "施工"],
'labelnames': ["护栏", "交通标志", "非交通标志", "施工锥桶", "施工水马"],
'trtFlag_seg': False,
'trtFlag_det': True,
'slopeIndex': [],
@ -574,14 +565,12 @@ class ModelType(Enum):
}},
'models':[
{
#'weight':'../weights/pth/AIlib2/channel2/yolov5.pt',
'weight':'../weights/trt/AIlib2/channel2/yolov5_%s_fp16.engine'%(gpuName),
'name':'yolov5',
'model':yolov5Model,
'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":{"0":0.7,"1":0.7,"2":0.8,"3":0.6} }
},
{
# 'weight' : '../weights/trt/AIlib2/ocr2/crnn_ch_4090_fp16_192X32.engine',
{
'weight' : '../weights/pth/AIlib2/ocr2/crnn_ch.pth',
'name':'ocr',
'model':ocrModel,
@ -596,7 +585,7 @@ class ModelType(Enum):
'std':[0.5,0.5,0.5],
'dynamic':False,
},
}
}
],
'detModelpara':[{"id":str(x),"config":{"k1":"v1","k2":"v2"}} for x in [0,1,2,3,4,5,6]],
'segPar': None,
@ -609,7 +598,7 @@ class ModelType(Enum):
},
'Segweights': None,
})
RIVERT_MODEL = ("25", "025", "河道检测模型(T)", 'riverT', lambda device, gpuName: {
'device': device,
'labelnames': ["漂浮物", "岸坡垃圾", "排口", "违建", "菜地", "水生植物", "河湖人员", "钓鱼人员", "船只",
@ -640,14 +629,10 @@ class ModelType(Enum):
"classes": 5,
"rainbows": COLOR
},
# "../weights/pth/AIlib2/%s/yolov5.pt" % modeType.value[3]
'Detweights': "../weights/trt/AIlib2/riverT/yolov5_%s_fp16.engine" % gpuName,
# '../weights/pth/AIlib2/%s/stdc_360X640.pth' % modeType.value[3]
'Segweights': '../weights/trt/AIlib2/riverT/stdc_360X640_%s_fp16.engine' % gpuName
})
FORESTCROWD_FARM_MODEL = ("26", "026", "森林人群模型", 'forestCrowd', lambda device, gpuName: {
'labelnames': ["林斑", "病死树", "行人", "火焰", "烟雾","人群"],
'postProcess':{'function':gather_post_process,'pars':{'pedestrianId':2,'crowdThreshold':4,'gatherId':5,'distancePersonScale':2.0}},
@ -678,7 +663,8 @@ class ModelType(Enum):
})
TRAFFICFORDSJ_FARM_MODEL = ("27", "027", "交通模型-大数据局", 'highWay2T', lambda device, gpuName: {
'device': str(device),
'labelnames': ["行人", "车辆", "纵向裂缝", "横向裂缝", "修补", "网状裂纹", "坑槽", "块状裂纹", "积水", "影子", "事故", "桥梁外观","设施破损缺失","龙门架","防抛网","标识牌损坏","护栏损坏","钢筋裸露" ],
'labelnames': ["行人", "车辆", "纵向裂缝", "横向裂缝", "修补", "网状裂纹", "坑槽", "块状裂纹", "积水", "影子",
"事故", "桥梁外观","设施破损缺失","龙门架","防抛网","标识牌损坏","护栏损坏","钢筋裸露" ],
'trtFlag_seg': True,
'trtFlag_det': True,
'seg_nclass': 3,
@ -694,12 +680,13 @@ class ModelType(Enum):
'function': tracfficAccidentMixFunction,
'pars': {
'modelSize': (640, 360),
#'modelSize': (1920,1080),
'RoadArea': 16000,
'roadVehicleAngle': 15,
'speedRoadVehicleAngleMax': 75,
'roundness': 1.0,
'cls': 10,
'CarId':1,
'CthcId':1,
'vehicleFactor': 0.1,
'confThres': 0.25,
'roadIou': 0.6,
@ -731,13 +718,13 @@ class ModelType(Enum):
'model':yolov5Model,
'par':{ 'half':True,'device':'cuda:0' ,'conf_thres':0.25,'iou_thres':0.45,'allowedList':list(range(20)),'segRegionCnt':1, 'trtFlag_det':True,'trtFlag_seg':False, "score_byClass":{"0":0.25,"1":0.3,"2":0.3,"3":0.3 } },
}
],
'postFile': {
"rainbows": COLOR
},
})
RUBBISH_MODEL = ("29", "029", "垃圾模型", 'rubbish', lambda device, gpuName: {
@ -751,15 +738,15 @@ class ModelType(Enum):
'model':yolov5Model,
'par':{ 'half':True,'device':'cuda:0' ,'conf_thres':0.25,'iou_thres':0.45,'allowedList':list(range(20)),'segRegionCnt':1, 'trtFlag_det':True,'trtFlag_seg':False, "score_byClass":{"0":0.25,"1":0.3,"2":0.3,"3":0.3 } },
}
],
'postFile': {
"rainbows": COLOR
},
})
FIREWORK_MODEL = ("30", "030", "烟花模型", 'firework', lambda device, gpuName: {
'labelnames': [ "烟花"],
'postProcess':{'function':default_mix,'pars':{}},
@ -771,13 +758,13 @@ class ModelType(Enum):
'model':yolov5Model,
'par':{ 'half':True,'device':'cuda:0' ,'conf_thres':0.25,'iou_thres':0.45,'allowedList':list(range(20)),'segRegionCnt':1, 'trtFlag_det':True,'trtFlag_seg':False, "score_byClass":{"0":0.25,"1":0.3,"2":0.3,"3":0.3 } },
}
],
'postFile': {
"rainbows": COLOR
},
})
TRAFFIC_SPILL_MODEL = ("50", "501", "高速公路抛洒物模型", 'highWaySpill', lambda device, gpuName: {
@ -785,26 +772,24 @@ class ModelType(Enum):
'labelnames': ["抛洒物","车辆"],
'trtFlag_seg': True,
'trtFlag_det': True,
'seg_nclass': 2,
'seg_nclass': 3,
'segRegionCnt': 2,
'segPar': {
#'modelSize': (640, 360),
'modelSize': (1920, 1080),
'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,
'mixFunction': {
'function': mixSpillage_postprocess,
'function': mixTraffic_postprocess,
'pars': {
#'modelSize': (640, 360),
'modelSize': (1920,1080),
'modelSize': (640, 360),
'RoadArea': 16000,
'roadVehicleAngle': 15,
'speedRoadVehicleAngleMax': 75,
'roundness': 1.0,
'cls': 1,
'cls': 0,
'vehicleFactor': 0.1,
'confThres': 0.25,
'roadIou': 0.6,
@ -832,26 +817,24 @@ class ModelType(Enum):
'labelnames': ["危化品","罐体","危险标识","普通车"],
'trtFlag_seg': True,
'trtFlag_det': True,
'seg_nclass': 2,
'seg_nclass': 3,
'segRegionCnt': 2,
'segPar': {
#'modelSize': (640, 360),
'modelSize': (1920, 1080),
'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,
'mixFunction': {
'function': mixCthc_postprocess,
'function': mixTraffic_postprocess,
'pars': {
#'modelSize': (640, 360),
'modelSize': (1920,1080),
'modelSize': (640, 360),
'RoadArea': 16000,
'roadVehicleAngle': 15,
'speedRoadVehicleAngleMax': 75,
'roundness': 1.0,
'cls': 4,
'cls': 0,
'vehicleFactor': 0.1,
'confThres': 0.25,
'roadIou': 0.6,
@ -865,7 +848,7 @@ class ModelType(Enum):
"name": "post_process",
"conf_thres": 0.25,
"iou_thres": 0.25,
"classes": 1,
"classes": 4,
"rainbows": COLOR
},
'detModelpara': [{"id": str(x), "config": {"k1": "v1", "k2": "v2"}} for x in [0]],
@ -896,34 +879,44 @@ class ModelType(Enum):
})
CITY_CARPLATE_MODEL = ("30", "301", "自研车牌检测", 'carplate', lambda device, gpuName: {
'labelnames': ["车牌"],
'device': str(device),
'models':{
'rainbows': COLOR,
'models': [
{
'weights': '../AIlib2/weights/conf/jkm/plate_yolov5s_v3.jit',
'conf_thres': 0.4,
'iou_thres': 0.45,
'nc':1,
'trtFlag_det': False,
'weight': '../weights/pth/AIlib2/carplate/plate_yolov5s_v3.jit',
'name': 'yolov5',
'model': yolov5Model,
'par': {
'device': 'cuda:0',
'half': False,
'conf_thres': 0.4,
'iou_thres': 0.45,
'nc': 1,
'plate_dilate': (0.5, 0.1)
},
},
{
'weight' : '../weights/pth/AIlib2/ocr2/crnn_ch.pth',
'name':'ocr',
'model':ocrModel,
'par':{
'char_file':'../AIlib2/conf/ocr2/benchmark.txt',
'mode':'ch',
'nc':3,
'imgH':32,
'imgW':192,
'hidden':256,
'mean':[0.5,0.5,0.5],
'std':[0.5,0.5,0.5],
'dynamic':False,
}
},
}
})
'trtFlag_ocr': False,
'weight': '../weights/pth/AIlib2/ocr2/crnn_ch.pth',
'name': 'ocr',
'model': ocrModel,
'par': {
'char_file': '../AIlib2/conf/ocr2/benchmark.txt',
'mode': 'ch',
'nc': 3,
'imgH': 32,
'imgW': 192,
'hidden': 256,
'mean': [0.5, 0.5, 0.5],
'std': [0.5, 0.5, 0.5],
'dynamic': False,
}
}],
})
CITY_INFRAREDPERSON_MODEL = ("30", "302", "红外行人模型", 'infraredperson', lambda device, gpuName: {
CITY_INFRAREDPERSON_MODEL = ("30", "302", "红外行人模型", 'infraredPerson', lambda device, gpuName: {
'labelnames': ["行人"],
'postProcess': {'function': default_mix, 'pars': {}},
'models':
@ -963,7 +956,87 @@ class ModelType(Enum):
"rainbows": COLOR
},
})
})
CITY_DENSECROWDCOUNT_MODEL = ("30", "304", "密集人群计数", 'DenseCrowdCount', lambda device, gpuName: {
'labelnames': ["人群计数"],
'device': str(device),
'rainbows': COLOR,
'models': [
{
'trtFlag_det': False,
'weight': "../weights/pth/AIlib2/DenseCrowd/SHTechA.pth", ###检测模型路径
'vggweight': "../weights/pth/AIlib2/DenseCrowd/vgg16_bn-6c64b313.pth", ###检测模型路径
'name': 'p2pnet',
'model': p2NnetModel,
'par': {
'device': 'cuda:0',
'row': 2,
'line': 2,
'point_loss_coef': 0.45,
'conf': 0.50,
'gpu_id': 0,
'eos_coef': '0.5',
'set_cost_class': 1,
'set_cost_point': 0.05,
'backbone': 'vgg16_bn'
},
}],
})
CITY_DENSECROWDESTIMATION_MODEL = ("30", "305", "密集人群密度估计", 'DenseCrowdEstimation', lambda device, gpuName: {
'labelnames': ["密度"],
'models':
[
{
'weight': "../weights/pth/AIlib2/DenseCrowd/SHTechA.pth", ###检测模型路径
'name': 'yolov5',
'model': yolov5Model,
'par': {'half': True, 'device': 'cuda:0', 'conf_thres': 0.50, 'iou_thres': 0.45,
'allowedList': list(range(20)), 'segRegionCnt': 1, 'trtFlag_det': True,
'trtFlag_seg': False, "score_byClass": {"0": 0.50, "1": 0.3, "2": 0.3, "3": 0.3}},
}
],
'postFile': {
"rainbows": COLOR
},
})
CITY_UNDERBUILDCOUNT_MODEL = ("30", "306", "建筑物下人群计数", 'perUnderBuild', lambda device, gpuName: {
'labelnames': ["建筑物下人群"],
'device': str(device),
'rainbows': COLOR,
'models': [
{
'weight': "../weights/trt/AIlib2/perUnderBuild/yolov5_%s_fp16.engine" % (gpuName), ###检测模型路径
'name': 'yolov5',
'model': yolov5Model,
'par': {'half': True, 'device': 'cuda:0', 'conf_thres': 0.25, 'iou_thres': 0.45,
'allowedList': [0,1,2], 'segRegionCnt': 1, 'trtFlag_det': True,
'trtFlag_seg': False, "score_byClass": {"0": 0.25, "1": 0.3, "2": 0.3, "3": 0.3}},
},
{
'trtFlag_det': False,
'weight': "../weights/pth/AIlib2/DenseCrowd/SHTechA.pth", ###检测模型路径
'vggweight': "../weights/pth/AIlib2/DenseCrowd/vgg16_bn-6c64b313.pth", ###检测模型路径
'name': 'p2pnet',
'model': p2NnetModel,
'par': {
'device': 'cuda:0',
'row': 2,
'line': 2,
'point_loss_coef': 0.45,
'conf': 0.50,
'gpu_id': 0,
'eos_coef': '0.5',
'set_cost_class': 1,
'set_cost_point': 0.05,
'backbone': 'vgg16_bn'
},
}],
})
@staticmethod
def checkCode(code):

116
util/ModelUtils.py
View File

@ -17,7 +17,7 @@ from util.PlotsUtils import get_label_arrays, get_label_array_dict
from util.TorchUtils import select_device
sys.path.extend(['..', '../AIlib2'])
from AI import AI_process, AI_process_forest, get_postProcess_para, ocr_process, AI_process_N, AI_process_C
from AI import AI_process, AI_process_forest, get_postProcess_para, ocr_process, AI_process_N, AI_process_C,AI_process_Ocr,AI_process_Crowd
from stdc import stdcModel
from segutils.segmodel import SegModel
from models.experimental import attempt_load
@ -46,8 +46,12 @@ class OneModel:
new_device = select_device(par.get('device'))
half = new_device.type != 'cpu'
Detweights = par['Detweights']
with open(Detweights, "rb") as f, trt.Runtime(trt.Logger(trt.Logger.ERROR)) as runtime:
model = runtime.deserialize_cuda_engine(f.read())
if par['trtFlag_det']:
with open(Detweights, "rb") as f, trt.Runtime(trt.Logger(trt.Logger.ERROR)) as runtime:
model = runtime.deserialize_cuda_engine(f.read())
else:
model = attempt_load(Detweights, map_location=new_device) # load FP32 model
if half: model.half()
par['segPar']['seg_nclass'] = par['seg_nclass']
Segweights = par['Segweights']
if Segweights:
@ -241,7 +245,7 @@ def channel2_process(args):
logger.error("算法模型分析异常：{}, requestId:{}", format_exc(), request_id)
raise ServiceException(ExceptionType.MODEL_ANALYSE_EXCEPTION.value[0],
ExceptionType.MODEL_ANALYSE_EXCEPTION.value[1])
def get_label_arraylist(*args):
width, height, names, rainbows = args
# line = int(round(0.002 * (height + width) / 2) + 1)
@ -354,6 +358,73 @@ def im_process(args):
raise ServiceException(ExceptionType.MODEL_ANALYSE_EXCEPTION.value[0],
ExceptionType.MODEL_ANALYSE_EXCEPTION.value[1])
def immulti_process(args):
model_conf, frame, requestId = args
device, modelList, detpar = model_conf[1], model_conf[2], model_conf[3]
try:
# new_device = torch.device(device)
# img, padInfos = pre_process(frame, new_device)
# pred = model(img)
# boxes = post_process(pred, padInfos, device, conf_thres=pardet['conf_thres'],
# iou_thres=pardet['iou_thres'], nc=pardet['nc']) # 后处理
return AI_process_Ocr([frame], modelList, device, detpar)
except ServiceException as s:
raise s
except Exception:
logger.error("算法模型分析异常：{}, requestId:{}", format_exc(), requestId)
raise ServiceException(ExceptionType.MODEL_ANALYSE_EXCEPTION.value[0],
ExceptionType.MODEL_ANALYSE_EXCEPTION.value[1])
class CARPLATEModel:
__slots__ = "model_conf"
def __init__(self, device, allowedList=None, requestId=None, modeType=None, gpu_name=None, base_dir=None,
env=None):
try:
logger.info("########################加载{}########################, requestId:{}", modeType.value[2],
requestId)
par = modeType.value[4](str(device), gpu_name)
modelList=[ modelPar['model'](weights=modelPar['weight'],par=modelPar['par']) for modelPar in par['models'] ]
detpar = par['models'][0]['par']
# new_device = torch.device(par['device'])
# modelList=[ modelPar['model'](weights=modelPar['weight'],par=modelPar['par']) for modelPar in par['models'] ]
logger.info("########################加载 plate_yolov5s_v3.jit 成功 ########################, requestId:{}",
requestId)
self.model_conf = (modeType, device, modelList, detpar, par['rainbows'])
except Exception:
logger.error("模型加载异常：{}, requestId:{}", format_exc(), requestId)
raise ServiceException(ExceptionType.MODEL_LOADING_EXCEPTION.value[0],
ExceptionType.MODEL_LOADING_EXCEPTION.value[1])
class DENSECROWDCOUNTModel:
__slots__ = "model_conf"
def __init__(self, device, allowedList=None, requestId=None, modeType=None, gpu_name=None, base_dir=None, env=None):
try:
logger.info("########################加载{}########################, requestId:{}", modeType.value[2],
requestId)
par = modeType.value[4](str(device), gpu_name)
rainbows = par["rainbows"]
models=[ modelPar['model'](weights=modelPar['weight'],par=modelPar['par']) for modelPar in par['models'] ]
postPar = [pp['par'] for pp in par['models']]
self.model_conf = (modeType, device, models, postPar, rainbows)
except Exception:
logger.error("模型加载异常：{}, requestId:{}", format_exc(), requestId)
raise ServiceException(ExceptionType.MODEL_LOADING_EXCEPTION.value[0],
ExceptionType.MODEL_LOADING_EXCEPTION.value[1])
def cc_process(args):
model_conf, frame, requestId = args
device, model, postPar = model_conf[1], model_conf[2], model_conf[3]
try:
return AI_process_Crowd([frame], model, device, postPar)
except ServiceException as s:
raise s
except Exception:
logger.error("算法模型分析异常：{}, requestId:{}", format_exc(), requestId)
raise ServiceException(ExceptionType.MODEL_ANALYSE_EXCEPTION.value[0],
ExceptionType.MODEL_ANALYSE_EXCEPTION.value[1])
# 百度AI图片识别模型
class BaiduAiImageModel:
@ -470,7 +541,7 @@ MODEL_CONFIG = {
lambda x, y, z: one_label(x, y, z),
lambda x: detSeg_demo2(x)
),
# 加载交通模型
ModelType.TRAFFIC_FARM_MODEL.value[1]: (
lambda x, y, r, t, z, h: OneModel(x, y, r, ModelType.TRAFFIC_FARM_MODEL, t, z, h),
@ -608,27 +679,27 @@ MODEL_CONFIG = {
),
# 加载交通模型
ModelType.TRAFFICFORDSJ_FARM_MODEL.value[1]: (
lambda x, y, r, t, z, h: OneModel(x, y, r, ModelType.TRAFFIC_FARM_MODEL, t, z, h),
ModelType.TRAFFIC_FARM_MODEL,
lambda x, y, r, t, z, h: OneModel(x, y, r, ModelType.TRAFFICFORDSJ_FARM_MODEL, t, z, h),
ModelType.TRAFFICFORDSJ_FARM_MODEL,
lambda x, y, z: one_label(x, y, z),
lambda x: model_process(x)
),
# 加载智慧工地模型
# 加载智慧工地模型
ModelType.SMARTSITE_MODEL.value[1]: (
lambda x, y, r, t, z, h: cityManagementModel(x, y, r, ModelType.SMARTSITE_MODEL, t, z, h),
ModelType.SMARTSITE_MODEL,
lambda x, y, z: one_label(x, y, z),
lambda x: detSeg_demo2(x)
),
# 加载垃圾模型
# 加载垃圾模型
ModelType.RUBBISH_MODEL.value[1]: (
lambda x, y, r, t, z, h: cityManagementModel(x, y, r, ModelType.RUBBISH_MODEL, t, z, h),
ModelType.RUBBISH_MODEL,
lambda x, y, z: one_label(x, y, z),
lambda x: detSeg_demo2(x)
),
# 加载烟花模型
ModelType.FIREWORK_MODEL.value[1]: (
lambda x, y, r, t, z, h: cityManagementModel(x, y, r, ModelType.FIREWORK_MODEL, t, z, h),
@ -657,6 +728,13 @@ MODEL_CONFIG = {
lambda x, y, z: one_label(x, y, z),
lambda x: detSeg_demo2(x)
),
# 加载自研车牌检测模型
ModelType.CITY_CARPLATE_MODEL.value[1]: (
lambda x, y, r, t, z, h: CARPLATEModel(x, y, r, ModelType.CITY_CARPLATE_MODEL, t, z, h),
ModelType.CITY_CARPLATE_MODEL,
None,
lambda x: immulti_process(x)
),
# 加载红外行人检测模型
ModelType.CITY_INFRAREDPERSON_MODEL.value[1]: (
lambda x, y, r, t, z, h: cityManagementModel(x, y, r, ModelType.CITY_INFRAREDPERSON_MODEL, t, z, h),
@ -670,5 +748,19 @@ MODEL_CONFIG = {
ModelType.CITY_NIGHTFIRESMOKE_MODEL,
lambda x, y, z: one_label(x, y, z),
lambda x: detSeg_demo2(x)
),
),
# 加载密集人群计数检测模型
ModelType.CITY_DENSECROWDCOUNT_MODEL.value[1]: (
lambda x, y, r, t, z, h: DENSECROWDCOUNTModel(x, y, r, ModelType.CITY_DENSECROWDCOUNT_MODEL, t, z, h),
ModelType.CITY_DENSECROWDCOUNT_MODEL,
None,
lambda x: cc_process(x)
),
# 加载建筑物下行人检测模型
ModelType.CITY_UNDERBUILDCOUNT_MODEL.value[1]: (
lambda x, y, r, t, z, h: DENSECROWDCOUNTModel(x, y, r, ModelType.CITY_UNDERBUILDCOUNT_MODEL, t, z, h),
ModelType.CITY_UNDERBUILDCOUNT_MODEL,
None,
lambda x: cc_process(x)
),
}

113
util/PlotsUtils.py
View File

@ -2,6 +2,7 @@ import cv2
import numpy as np
from PIL import Image, ImageDraw, ImageFont
import unicodedata
from loguru import logger
FONT_PATH = "../AIlib2/conf/platech.ttf"
zhFont = ImageFont.truetype(FONT_PATH, 20, encoding="utf-8")
@ -67,6 +68,12 @@ def xywh2xyxy2(param):
# return [(lt, yt), (rt, yt), (rt, yr), (lt, yr)]
return np.asarray(param[0][0:4], np.int32), float(param[1]), int(param[2])
def xy2xyxy(box):
if not isinstance(box[0], (list, tuple, np.ndarray)):
x1, y1, x2, y2 = int(box[0]), int(box[1]), int(box[2]), int(box[3])
# 顺时针
box = [(x1, y1), (x2, y1), (x2, y2), (x1, y2)]
return box
def draw_painting_joint(box, img, label_array, score=0.5, color=None, config=None, isNew=False):
# 识别问题描述图片的高、宽
@ -218,6 +225,11 @@ def draw_name_joint(box, img, label_array_dict, score=0.5, color=None, config=No
cv2.putText(img, label, p3, 0, config[3], [225, 255, 255], thickness=config[4], lineType=cv2.LINE_AA)
return img, box
def draw_name_ocr(box, img, color, line_thickness=2, outfontsize=40):
font = ImageFont.truetype(FONT_PATH, outfontsize, encoding='utf-8')
# (color=None, label=None, font=None, fontSize=40, unify=False)
label_zh = get_label_array(color, box[0], font, outfontsize)
return plot_one_box_auto(box[1], img, color, line_thickness, label_zh)
def filterBox(det0, det1, pix_dis):
# det0为 (m1, 11) 矩阵
@ -251,8 +263,105 @@ def filterBox(det0, det1, pix_dis):
x_c, y_c = (x3+x4)//2, (y3+y4)//2
dis = (x2_c - x_c)**2 + (y2_c - y_c)**2
mask = (joint_det[..., 9] == joint_det[..., 20]) & (dis <= pix_dis**2)
# 类别相同 & 中心点在上一帧的框内 判断为True
res = np.sum(mask, axis=1)
det0_copy[..., -1] = res
return det0_copy
return det0_copy
def plot_one_box_auto(box, img, color=None, line_thickness=2, label_array=None):
# print("省略 :%s, box:%s"%('+++' * 10, box))
# 识别问题描述图片的高、宽
lh, lw = label_array.shape[0:2]
# print("省略 :%s, lh:%s, lw:%s"%('+++' * 10, lh, lw))
# 图片的长度和宽度
imh, imw = img.shape[0:2]
box = xy2xyxy(box)
# 框框左上的位置
x0, y1 = box[0][0], box[0][1]
# print("省略 :%s, x0:%s, y1:%s"%('+++' * 10, x0, y1))
x1, y0 = x0 + lw, y1 - lh
# 如果y0小于0, 说明超过上边框
if y0 < 0:
y0 = 0
# y1等于文字高度
y1 = y0 + lh
# 如果y1框框的高大于图片高度
if y1 > imh:
# y1等于图片高度
y1 = imh
# y0等于y1减去文字高度
y0 = y1 - lh
# 如果x0小于0
if x0 < 0:
x0 = 0
x1 = x0 + lw
if x1 > imw:
x1 = imw
x0 = x1 - lw
# box_tl = max(int(round(imw / 1920 * 3)), 1) or round(0.002 * (imh + imw) / 2) + 1
'''
1. img（array） 为ndarray类型（可以为cv.imread）直接读取的数据
2. box（array）：为所画多边形的顶点坐标
3. 所画四边形是否闭合，通常为True
4. color（tuple）：BGR三个通道的值
5. thickness（int）：画线的粗细
6. shift：顶点坐标中小数的位数
'''
# Plots one bounding box on image img
tl = line_thickness or round(0.002 * (img.shape[0] + img.shape[1]) / 2) + 1 # line/font thickness
box1 = np.asarray(box, np.int32)
cv2.polylines(img, [box1], True, color, tl)
img[y0:y1, x0:x1, :] = label_array
return img, box
def draw_name_crowd(dets, img, color, line_thickness=2, outfontsize=20):
font = ImageFont.truetype(FONT_PATH, outfontsize, encoding='utf-8')
if len(dets) == 1:
label = '当前人数：%d'%len(dets[0])
detP = dets[0]
for p in detP:
img = cv2.circle(img, (int(p[0]), int(p[1])), line_thickness, color, -1)
label_arr = get_label_array(color, label, font, outfontsize)
lh, lw = label_arr.shape[0:2]
img[0:lh, 0:lw, :] = label_arr
elif len(dets) == 2:
detP = dets[1]
for p in detP:
img = cv2.circle(img, (int(p[0]), int(p[1])), line_thickness, color, -1)
detM = dets[0]
h, w = img.shape[:2]
for b in detM:
label = '该建筑下行人及数量：%d'%(int(b[4]))
label_arr = get_label_array(color, label, font, outfontsize)
lh, lw = label_arr.shape[0:2]
# 框框左上的位置
x0, y1 = int(b[0]), int(b[1])
# print("省略 :%s, x0:%s, y1:%s"%('+++' * 10, x0, y1))
x1, y0 = x0 + lw, y1 - lh
# 如果y0小于0, 说明超过上边框
if y0 < 0:
y0 = 0
# y1等于文字高度
y1 = y0 + lh
# 如果y1框框的高大于图片高度
if y1 > h:
# y1等于图片高度
y1 = h
# y0等于y1减去文字高度
y0 = y1 - lh
# 如果x0小于0
if x0 < 0:
x0 = 0
x1 = x0 + lw
if x1 > w:
x1 = w
x0 = x1 - lw
cv2.polylines(img, [np.asarray(xy2xyxy(b), np.int32)], True, (0, 128, 255), 2)
img[y0:y1, x0:x1, :] = label_arr
return img, dets