🐛 2021年7月4日16:08:12 修复预训练模型会下载失败的问题

ExcaliburEX · Jul 4, 2021 · 66d9c9f · 66d9c9f
1 parent d414dcd
commit 66d9c9f
Show file tree

Hide file tree

Showing 2 changed files with 108 additions and 12 deletions.
diff --git a/README.md b/README.md
@@ -4,8 +4,10 @@
 ---
 
 最新release：
-<h3 align="center"><a href= 'https://github.com/ExcaliburEX/FingerveinRecognition/releases/download/V1.0/FingerveinRecogntion.exe'><img alt="GitHub release (latest by date)" src="https://img.shields.io/github/downloads/ExcaliburEX/FingerveinRecogntion/V1.0/total?color=success&flat-square&logo=Cachet"></a></h3>
+<h3 align="center"><a href= 'https://github.com/ExcaliburEX/FingerveinRecognition/releases/download/V1.1/FingerveinRecogntion.exe'><img alt="GitHub release (latest by date)" src="https://img.shields.io/github/downloads/ExcaliburEX/FingerveinRecognition/V1.1/total?color=success&flat-square&logo=Cachet"></a></h3>
 
+# 更新日志
+- 2021-07-04：因为VGG19下载的预训练模型在Github导致有可能下载失败，因此将模型文件存储到自己的腾讯云OSS桶，通过程序自动下载，手动设置了VGG19的预训练模型的路径，规避了因为网络问题导致的下载失败。
 
 # 使用说明
 

diff --git a/finger.py b/finger.py
@@ -8,12 +8,12 @@
 from tensorflow.compat.v1 import logging as log
 log.set_verbosity(log.ERROR)
 import time
-from keras.applications.vgg16 import VGG16, preprocess_input
+from keras.applications.vgg19 import VGG19, preprocess_input
 from keras.preprocessing import image
 from PIL import Image
 import numpy as np
 import random
-import ssl
+# import ssl
 
 
 def get_feature(path,model):
@@ -38,6 +38,7 @@ def cos_sim(a, b):
 import PySimpleGUI as sg
 from io import BytesIO
 import base64
+import requests
 
 def gaborfilter(I, S, F, W, P):
     size = int(1.5 / S)
@@ -133,7 +134,7 @@ def Train(file,flag,window,model):
             f.close()
         return feature,1
     else:
-        print('\n',time.strftime("[%Y-%m-%d %H:%M:%S]: ",time.localtime()), "图片 %s 已识别过..."%(name))
+        print(time.strftime("\n[%Y-%m-%d %H:%M:%S]: ",time.localtime()), "图片 %s 已识别过..."%(name))
     return None,2
 
 def Test(trainPic, readPic, finger_num,num,window,model):
@@ -220,18 +221,22 @@ def GUI():
     layout = [
         [
             sg.Text('单人手指图片个数：', font=("KaiTi", 12),justification='left',relief=sg.RELIEF_RIDGE),
+            # sg.InputText('7', font=("KaiTi", 12), size=(17, 2), key='-RANGE1-'),
             sg.InputText('', font=("KaiTi", 12), size=(17, 2), key='-RANGE1-'),
             sg.Text('随机选取的图片测试个数：', font=("KaiTi", 12),justification='left',relief=sg.RELIEF_RIDGE),
+            # sg.InputText('5', font=("KaiTi", 12), size=(17, 2), key='-RANGE2-')
             sg.InputText('', font=("KaiTi", 12), size=(17, 2), key='-RANGE2-')
             ],
         [
             sg.Text('训练集图片文件夹：', font=("KaiTi", 12),justification='left',relief=sg.RELIEF_RIDGE),
-            sg.In(size=(60, 1), enable_events=True, key="-FOLDER1-"),
+            # sg.In('C:/Users/Excalibur/Desktop/Test/V1.0/HighGuardFinger',size=(60, 1), enable_events=True, key="-FOLDER1-"),
+            sg.In('',size=(60, 1), enable_events=True, key="-FOLDER1-"),
             sg.FolderBrowse('浏览', button_color=('Lavender', 'BlueViolet'), font=("KaiTi", 10),size=(8, 1))
         ],
         [
             sg.Text('待识别图片文件夹：', font=("KaiTi", 12),justification='left',relief=sg.RELIEF_RIDGE),
-            sg.In(size=(60, 1), enable_events=True, key="-FOLDER2-"),
+            # sg.In('C:/Users/Excalibur/Desktop/Test/V1.0/HighGuardTest',size=(60, 1), enable_events=True, key="-FOLDER2-"),
+            sg.In('',size=(60, 1), enable_events=True, key="-FOLDER2-"),
             sg.FolderBrowse('浏览', button_color=('Lavender', 'BlueViolet'), font=("KaiTi", 10),size=(8, 1))
         ],
         [sg.Column(col, background_color='papayawhip', size=(
@@ -244,7 +249,7 @@ def GUI():
             sg.Exit('退出', button_color=('white', 'firebrick4'),key='Exit', size=(37, 1), font=("KaiTi", 12))
         ]
     ]
-    window = sg.Window('HighGuardFingerRecV1.0', layout,
+    window = sg.Window('HighGuardFingerRecV1.1', layout,
             default_element_size=(80, 1), resizable=True, element_justification='center', text_justification='center',finalize=True)
 
 
@@ -269,30 +274,119 @@ def GUI():
 
 
 
+# 判斷模型文件是否存在
+def FileExist():
+    filename = 'vgg19_weights_tf_dim_ordering_tf_kernels_notop.h5'
+    # username = getpass.getuser()
+    # fileloc = r'C:\Users\\' + username + '\.keras\models\\'
+    fileloc = os.getcwd() + '\\'
+    fullname = fileloc + filename
+    if os.path.exists(fullname):
+        return True,fileloc
+    else:
+        return False,fileloc
+
+# 手动下载模型文件
+def progressbar(url,path,file_name,window):
+    if not os.path.exists(path):   # 看是否有该文件夹，没有则创建文件夹
+         os.mkdir(path)
+    file_path = os.path.join(path, file_name)
+    start = time.time() #下载开始时间
+    response = requests.get(url, stream=True)
+    size = 0    #初始化已下载大小
+    chunk_size = 1024 * 1024 * 3  # 每次下载的数据大小
+    content_size = int(response.headers['content-length'])  # 下载文件总大小
+    count_tmp = 0
+    count = 0
+    try:
+        if response.status_code == 200:   #判断是否响应成功
+            print(time.strftime("[%Y-%m-%d %H:%M:%S]: ",time.localtime()),'开始下载,[模型-大小] %s : %.2f MB'%(file_name,content_size / chunk_size * 3))   #开始下载，显示下载文件大小
+            with open(file_name,'wb') as file:   #显示进度条
+                for data in response.iter_content(chunk_size = chunk_size):
+                    file.write(data)
+                    count += len(data)
+                    size += len(data)
+                    speed = (count - count_tmp) / 1024 / 1024
+                    count_tmp = count
+                    end = time.time()   #下载结束时间
+                    rate = int((size / content_size) * 20)
+                    len1 = '>' * rate
+                    len2 = '_' * (20-rate)
+                    window['OUTPUT'].update(value=time.strftime("[%Y-%m-%d %H:%M:%S]: ",time.localtime()) + '模型总大小：%.2f MB，下载速度：%.2f M/S \n'%(content_size / chunk_size * 3,speed) + 
+                    time.strftime("[%Y-%m-%d %H:%M:%S]: ",time.localtime()) + '下载进度: [%s%s] |%d / %d| %.2f%%' % (len1,len2,size,content_size, float(size / content_size * 100)) +
+                    time.strftime("\n[%Y-%m-%d %H:%M:%S]: ",time.localtime()) + '已经耗时%.2f秒'%(end - start))
+                    time.sleep(0.1)
+        end = time.time()   #下载结束时间
+        print(time.strftime("\n[%Y-%m-%d %H:%M:%S]: ",time.localtime()),'模型下载成功!,耗时: %.2f秒' % (end - start))  #输出下载用时时间
+        return True
+    except:
+        end = time.time()   #下载结束时间
+        print(time.strftime("[%Y-%m-%d %H:%M:%S]: ",time.localtime()),'模型下载失败!,耗时: %.2f秒，请检查网络及相关设置再继续训练' % (end - start))  #输出下载用时时间
+        return False
+
+
+# 暂时放弃的第一版下载方案
+# def DownloadFile2(model_url, save_url,file_name):
+#     try:
+#         if model_url is None or save_url is None or file_name is None:
+#             print('参数错误')
+#             return None
+#         folder = os.path.exists(save_url)
+#         if not folder:
+#             os.makedirs(save_url)
+#         res = requests.get(model_url,stream=True) 
+#         total_size = int(int(res.headers["Content-Length"])/1024+0.5)
+#         file_path = os.path.join(save_url, file_name)
+#         from tqdm import tqdm
+#         with open(file_path, 'wb') as fd:
+#             print('开始下载文件：{},当前文件大小：{}KB'.format(file_name,total_size))
+#             for chunk in tqdm(iterable=res.iter_content(1024),total=total_size,unit='k',desc=None):
+#                 fd.write(chunk)
+#             print(file_name+' 下载完成！')
+#     except:
+#         print("程序错误")
+
+
+
+
+
+
 # 批量训练与识别
 def BatchTrain(directory,window):
+    model_name = 'VGG19'
+    model_full_name = 'vgg19_weights_tf_dim_ordering_tf_kernels_notop.h5'
     print(time.strftime("[%Y-%m-%d %H:%M:%S]: ",time.localtime()), "正在加载识别模型..." )
-    ssl._create_default_https_context = ssl._create_unverified_context # 下载模型的时候不想进行ssl证书校验
+    # ssl._create_default_https_context = ssl._create_unverified_context # 下载模型的时候不想进行ssl证书校验
     starttime = time.time()
-    model = VGG16(weights='imagenet', include_top=False)
+    file_name = 'vgg19_weights_tf_dim_ordering_tf_kernels_notop.h5'
+    model_url = 'https://blog-1259799643.cos.ap-shanghai.myqcloud.com/' + file_name
+    flag,loc = FileExist()
+    if flag == False:
+        print(time.strftime("[%Y-%m-%d %H:%M:%S]: ",time.localtime()), "初次使用需要下载 %s 预训练模型！"%(model_name))
+        downloadflag = progressbar(model_url, loc, file_name,window)
+        if downloadflag == False:
+            return
+    model = VGG19(weights=os.getcwd() + '\\' + model_full_name, include_top=False)
     elapse = time.time() - starttime
     print(time.strftime("[%Y-%m-%d %H:%M:%S]: ",time.localtime()), "加载完毕，耗时：%.3fs" % (elapse))
     filelist = os.listdir(directory)
     cnt = 1
     starttime2 = time.time()
     for f in filelist:
-        window['OUTPUT'].update(value=time.strftime("[%Y-%m-%d %H:%M:%S]: ",time.localtime()) +  "正在训练第 %d / %d 个图片..."%(cnt,len(filelist)))
+        elapse2 = time.time() - starttime2
+        window['OUTPUT'].update(value=time.strftime("[%Y-%m-%d %H:%M:%S]: ",time.localtime()) +  "正在训练第 %d / %d 个图片： %s，已耗时：%.2fs，单张平均耗时：%.2fs ..."%(cnt,len(filelist),f,elapse2,elapse2/cnt))
         Train(directory + '/' + f,1,window,model)
         cnt += 1
     elapse2 = time.time() - starttime2
-    print('\n',time.strftime("[%Y-%m-%d %H:%M:%S]: ",time.localtime()), "训练完毕，耗时：%.3fs" % (elapse2))
+    print(time.strftime("\n[%Y-%m-%d %H:%M:%S]: ",time.localtime()), "训练完毕，耗时：%.3fs" % (elapse2))
 
 
 def BatchTest(trainPic, readPicDir, finger_num,num,window):
+    model_full_name = 'vgg19_weights_tf_dim_ordering_tf_kernels_notop.h5'
     print(time.strftime("[%Y-%m-%d %H:%M:%S]: ",time.localtime()), "单人手指数：%d张，随机抽取其中 %d 张进行比对！"%(int(finger_num),int(num)) )
     print(time.strftime("[%Y-%m-%d %H:%M:%S]: ",time.localtime()), "正在加载识别模型..." )
     starttime = time.time()
-    model = VGG16(weights='imagenet', include_top=False)
+    model = VGG19(weights=os.getcwd() + '\\' + model_full_name, include_top=False)
     elapse = time.time() - starttime
     print(time.strftime("[%Y-%m-%d %H:%M:%S]: ",time.localtime()), "加载完毕，耗时：%.3fs" % (elapse))
     filelist = os.listdir(readPicDir)