import re
from django.shortcuts import render
from django.http import HttpResponse, JsonResponse
import os
import subprocess
from datetime import datetime
import learn.yolov5.detect as DL
import learn.Darknet2Pytorch2Onnx.hy_runonnx as RONNX
from pathlib import PosixPath
import requests, os, json
import cv2
import numpy as np
def adb_shell(cmd):
result = subprocess.getstatusoutput(cmd)
return result
'''改程序执行慢，且因为C++和python在cv2中处理细节不同，返回值并不相同
def detect_brightness():
img = cv2.imread("/home/wangchunlin/img_brightness/1.jpeg")
gray_img= cv2.cvtColor(img, cv2.COLOR_RGB2GRAY)
a=0
Ma = 0
hist = np.zeros(256,dtype=int)
print(hist.dtype)
h, w = gray_img.shape[:2]
print(h, w)
for i in range(h):
for j in range(w):
if(float(gray_img[i,j]-128)>128):
input(">")
a+=float(gray_img[i,j]-128)
hist[int(gray_img[i,j])]+=1
da = a /float(h*w)
print(da)
for i in range(256):
Ma+=abs(i-128-da)*hist[i]
Ma /= float(h*w)
print(Ma)
cast=abs(da)/abs(Ma)
return cast
'''
def upload(request):
if request.method == 'POST':
file = request.FILES.get('file') #获取前端上传的文件
m_model=request.POST.get('model')
m_class=request.POST.get('class')
model_name = request.POST.get("model_name")
class_name = request.POST.get("class_name")
print(request.POST["model_name"])
print(request.POST["class_name"])
fix = datetime.now().strftime('%Y%m%d%H%M%S%f')+'1' #给文件加前缀防止文件名重复
#以下用绝对路径存储文件，之前我用相对路径一直写不对
curPath = os.path.abspath( os.path.dirname( __file__ ) )
img_path = os.path.abspath(curPath+'/static/upload/'+fix+file.name)
#返回给前端的图片路径用相对路径，前端用绝对路径反而加载不了图片
img_path_res = '/static/detected/'+fix+file.name
print(img_path)
f = open(img_path,'wb')
for i in file.chunks():
f.write(i)
f.close()
return inference(m_class, m_model, curPath, img_path_res, img_path, model_name, class_name)
def dlurl(request):
if request.method == 'GET':
url = request.GET["url"] #获取前端上传的文件
m_model=request.GET["model"]
m_class=request.GET["class"]
model_name = request.GET.get("model_name")
class_name = request.GET.get("class_name")
print(request.GET["model_name"])
print(request.GET["class_name"])
print("m_model:",m_model)
print("url:",url)
name = url.split(r'/')[-1]
if not (name.endswith(".jpg") or name.endswith(".jpeg") or name.endswith(".png") or name.endswith(".bmp")):
return JsonResponse({'flag': False})
fix = datetime.now().strftime('%Y%m%d%H%M%S%f')+'1' #给文件加前缀防止文件名重复
curPath = os.path.abspath( os.path.dirname( __file__ ) )
img_path = os.path.abspath(curPath+'/static/download/'+fix+name)
r = requests.get(url)
# 保存
with open (img_path, 'wb') as f:
f.write(r.content)
f.close
#返回给前端的图片路径用相对路径，前端用绝对路径反而加载不了图片
img_path_res = '/static/detected/'+fix+name
return inference(m_class, m_model, curPath, img_path_res, img_path, model_name, class_name)
'''
result = adb_shell("ImageDetect/build/ImageDetect {}".format(img_path))
if result[0]==0:
cast = result[1].split("||")[0]
da = result[1].split("||")[1]
satu = result[1].split("||")[2]
CON = result[1].split("||")[3]
ASM = result[1].split("||")[4]
ENT = result[1].split("||")[5]
MEAN = result[1].split("||")[6]
sb = result[1].split("||")[7]
lap = result[1].split("||")[8]
dev = result[1].split("||")[9]
print(cast, da, satu)
mesg = "颜色分布(0-1合理):{} 分布方差(负数偏暗):{} 色彩饱和度(零为补光):{}
对比度:{} 角度方向二阶矩:{} 熵:{} 平均值:{} Sobel:{} Lap:{} dev:{}".format(cast, da, satu, CON, ASM, ENT, MEAN,sb,lap,dev)
if(int(m_class)==1):
if(int(m_model)==1): #小模型
RR = DL.run(weights=(curPath+"/yolov5/v3s.pt"), source=img_path, project=(curPath+"/static/detected"))
return JsonResponse({'img_name':img_path_res,'code':"【烟火 小模型】"+'
'+RR+'
'+mesg, 'flag': True})
if(int(m_model)==2): #大模型
RR = DL.run(weights=(curPath+"/yolov5/v6m6.pt"), source=img_path, imgsz=(1280, 1280), project=(curPath+"/static/detected"))
return JsonResponse({'img_name':img_path_res,'code':"【烟火 大模型】"+'
'+RR+'
'+mesg, 'flag': True})
if(int(m_model)==3): #泛模型
RR = DL.run(weights=(curPath+"/yolov5/best.pt"), source=img_path, project=(curPath+"/static/detected"))
return JsonResponse({'img_name':img_path_res,'code':"【烟火 泛模型】"+'
'+RR+'
'+mesg, 'flag': True})
if(int(m_class)==2):
RR = DL.run(weights=(curPath+"/yolov5/yolov5s.pt"), source=img_path, project=(curPath+"/static/detected"))
return JsonResponse({'img_name':img_path_res,'code':"【一般物体检测】"+'
'+RR+'
'+mesg, 'flag': True})
if(int(m_class)==3):
RR = DL.run(weights=(curPath+"/yolov5/fire_smoke_towercrane.pt"), source=img_path, project=(curPath+"/static/detected"))
return JsonResponse({'img_name':img_path_res,'code':"【烟火+塔吊】"+'
'+RR+'
'+mesg, 'flag': True})
if(int(m_class)==4):
RR = DL.run(weights=(curPath+"/yolov5/vehicle_4class_ljp_v3.0.pt"), source=img_path, project=(curPath+"/static/detected"))
return JsonResponse({'img_name':img_path_res,'code':"【外破四类】"+'
'+RR+'
'+mesg, 'flag': True})
if(int(m_class)==5):
RR = DL.run(weights=(curPath+"/yolov5/vehicle_4class_ljp_v7.0.pt"), source=img_path, project=(curPath+"/static/detected"))
return JsonResponse({'img_name':img_path_res,'code':"【外破四类】"+'
'+RR+'
'+mesg, 'flag': True})
if(int(m_class)==6):
RR = DL.run(weights=(curPath+"/yolov5/vehicle_4class_ljp_v7.0m.pt"), source=img_path, project=(curPath+"/static/detected"))
return JsonResponse({'img_name':img_path_res,'code':"【外破四类】"+'
'+RR+'
'+mesg, 'flag': True})
if(int(m_class)==7):
RR = DL.run(weights=(curPath+"/yolov5/birdpecked_unnormal_version1.pt"), source=img_path, project=(curPath+"/static/detected"))
return JsonResponse({'img_name':img_path_res,'code':"【绝缘子两类】"+'
'+RR+'
'+mesg, 'flag': True})
'''
def home(request):
return render(request, 'home.html')
def inference(m_class, m_model, curPath, img_path_res, img_path, model_name, class_name):
result = adb_shell("ImageDetect/build/ImageDetect {}".format(img_path))
if result[0]==0:
cast = result[1].split("||")[0]
da = result[1].split("||")[1]
satu = result[1].split("||")[2]
CON = result[1].split("||")[3]
ASM = result[1].split("||")[4]
ENT = result[1].split("||")[5]
MEAN = result[1].split("||")[6]
sb = result[1].split("||")[7]
lap = result[1].split("||")[8]
dev = result[1].split("||")[9]
print(cast, da, satu)
mesg = "颜色分布(0-1合理):{} 分布方差(负数偏暗):{} 色彩饱和度(零为补光):{}
对比度:{} 角度方向二阶矩:{} 熵:{} 平均值:{} Sobel:{} Lap:{} dev:{}".format(cast, da, satu, CON, ASM, ENT, MEAN,sb,lap,dev)
if(int(m_class)==1): #外破
if(int(m_model)==1):
RR = DL.run(weights=(curPath+"/yolov5/vehicle_4class_ljp_v3.0.pt"), source=img_path, project=(curPath+"/static/detected"))
return JsonResponse({'img_name':img_path_res,'code':class_name+":"+model_name+",模型: "+"vehicle_4class_ljp_v3.0.pt"+'
'+RR+'
'+mesg, 'flag': True})
if(int(m_model)==2):
RR = DL.run(weights=(curPath+"/yolov5/vehicle_4class_ljp_v7.0.pt"), source=img_path, project=(curPath+"/static/detected"))
return JsonResponse({'img_name':img_path_res,'code':class_name+":"+model_name+",模型: "+"vehicle_4class_ljp_v7.0.pt"+'
'+RR+'
'+mesg, 'flag': True})
if(int(m_model)==3):
RR = DL.run(weights=(curPath+"/yolov5/vehicle_4class_ljp_v7.0m.pt"), source=img_path, project=(curPath+"/static/detected"))
return JsonResponse({'img_name':img_path_res,'code':class_name+":"+model_name+",模型: "+"vehicle_4class_ljp_v7.0m.pt"+'
'+RR+'
'+mesg, 'flag': True})
if(int(m_model)==4):
RR = DL.run(weights=(curPath+"/yolov5/fire_smoke_towercrane.pt"), source=img_path, project=(curPath+"/static/detected"))
return JsonResponse({'img_name':img_path_res,'code':class_name+":"+model_name+",模型: "+"fire_smoke_towercrane.pt"+'
'+RR+'
'+mesg, 'flag': True})
if(int(m_model)==5):
RR = DL.run(weights=(curPath+"/yolov5/vehicle_4class_ljp_v7.0m6.pt"), source=img_path, project=(curPath+"/static/detected"),imgsz=(1280,1280))
return JsonResponse({'img_name':img_path_res,'code':class_name+":"+model_name+",模型: "+"vehicle_4class_ljp_v7.0m6.pt"+'
'+RR+'
'+mesg, 'flag': True})
if(int(m_model)==6):
RR=RONNX.main(curPath+"/Darknet2Pytorch2Onnx/hy_vehicle_4class.onnx", curPath+"/Darknet2Pytorch2Onnx/hy_vehicle_4class.names", img_path, (curPath+"/static/detected"))
return JsonResponse({'img_name':img_path_res,'code':class_name+":"+model_name+",模型: "+"hy_vehicle_4class.onnx"+'
'+RR+'
'+mesg, 'flag': True})
if(int(m_model)==7):
RR=RONNX.main(curPath+"/Darknet2Pytorch2Onnx/hy_vehicle_2class.onnx", curPath+"/Darknet2Pytorch2Onnx/hy_vehicle_2class.names", img_path, (curPath+"/static/detected"))
return JsonResponse({'img_name':img_path_res,'code':class_name+":"+model_name+",模型: "+"hy_vehicle_2class.onnx"+'
'+RR+'
'+mesg, 'flag': True})
if(int(m_class)==2): #山火
if(int(m_model)==1): # 小
RR = DL.run(weights=(curPath+"/yolov5/v3s.pt"), source=img_path, project=(curPath+"/static/detected"))
return JsonResponse({'img_name':img_path_res,'code':class_name+":"+model_name+",模型: "+"v3s.pt"+'
'+RR+'
'+mesg, 'flag': True})
if(int(m_model)==2): # 大
RR = DL.run(weights=(curPath+"/yolov5/v6m6.pt"), source=img_path, imgsz=(1280, 1280), project=(curPath+"/static/detected"))
return JsonResponse({'img_name':img_path_res,'code':class_name+":"+model_name+",模型: "+"v6m6.pt"+'
'+RR+'
'+mesg, 'flag': True})
if(int(m_model)==3): # 范模型
RR = DL.run(weights=(curPath+"/yolov5/best.pt"), source=img_path, project=(curPath+"/static/detected"))
return JsonResponse({'img_name':img_path_res,'code':class_name+":"+model_name+",模型: "+"best.pt"+'
'+RR+'
'+mesg, 'flag': True})
if(int(m_class)==3): # 绝缘子
if(int(m_model)==1):
RR = DL.run(weights=(curPath+"/yolov5/birdpecked_unnormal_version1.pt"), source=img_path, project=(curPath+"/static/detected"))
return JsonResponse({'img_name':img_path_res,'code':class_name+":"+model_name+",模型: "+"birdpecked_unnormal_version1.pt"+'
'+RR+'
'+mesg, 'flag': True})
if(int(m_class)==4): # 一般物体coco
if(int(m_model)==1): #yolov5s
RR = DL.run(weights=(curPath+"/yolov5/yolov5s.pt"), source=img_path, project=(curPath+"/static/detected"))
return JsonResponse({'img_name':img_path_res,'code':class_name+":"+model_name+",模型: "+"yolov5s.pt"+'
'+RR+'
'+mesg, 'flag': True})