bug fix

lib/common.py

@@ -8,7 +8,7 @@
 
 def print_start():
     """
-    打印 开始 logog
+    打印 开始 logo
     :return:
     """
 

lib/gen/generator.py

@@ -13,17 +13,19 @@
 from lib.gen.static import *
 
 # 路径先采用绝对路径
+test_gen_path = '/Volumes/doc/projects/ml/vin_recognize/static/test/'
 images_path = "/Volumes/doc/projects/ml/vin_recognize/static/images/"
 font_path = '/Volumes/doc/projects/ml/vin_recognize/static/font/'
 bad_path = '/Volumes/doc/projects/ml/vin_recognize/static/bad/'
 predict_path = '/Volumes/doc/projects/ml/vin_recognize/static/predict/'
 font = 'OCR-B.ttf'
 
+
 class GenVin:
 
-    def __init__(self, height, width,predict_path=predict_path,fontSize=30,
-                 NoPlates=bad_path,font_path=font_path,
-                 images_path=images_path,font=font):
+    def __init__(self, height, width, predict_path=predict_path, fontSize=30,
+                 NoPlates=bad_path, font_path=font_path,
+                 images_path=images_path, font=font):
 
         self.fontSize = fontSize
         self.font = ImageFont.truetype(font_path + font, self.fontSize, 0)
@@ -56,11 +58,12 @@ def draw(self, val):
         :param val:
         :return:
         """
-        height,width =  self.bg.shape
+        height, width = self.bg.shape
         offset = 0
-        base = int((width - self.max_size*self.fontSize)/2)
+        base = int((width - self.max_size * self.fontSize) / 2)
         for i, item in enumerate(val):
-            self.bg[int((height-self.fontSize)/2):int((height+self.fontSize)/2), base: base + self.fontSize] = self.draw_item(item, 0, 0)
+            self.bg[int((height - self.fontSize) / 2):int((height + self.fontSize) / 2),
+            base: base + self.fontSize] = self.draw_item(item, 0, 0)
             base += self.fontSize + offset
         return self.bg
 
@@ -86,13 +89,13 @@ def GenCh(self, val, img):
         """
         img = Image.new("RGB", (self.fontSize, self.bg.shape[1]), (0, 0, 0))
         draw = ImageDraw.Draw(img)
-        draw.text((0, int(self.bg.shape[1]-self.fontSize)/2), val, (255, 255), font=self.font)
+        draw.text((0, int(self.bg.shape[1] - self.fontSize) / 2), val, (255, 255), font=self.font)
         A = np.array(img)
         return A
 
-    def random_start(self,seed):
+    def random_start(self, seed):
         ran = r(seed)
-        return ran%3 == 0
+        return ran % 3 == 0
 
     def generate(self, text):
         fg = self.draw(text)
@@ -125,7 +128,7 @@ def gen_sample(self):
         return label, img[:, :, 2]  # 返回的label为标签，img为深度为3的图像像素
 
     def gen_img(self):
-        text, label = self.get_text_v2()
+        text, label = self.get_text()
         img = self.generate(text)
         img = cv2.resize(img, (self.width, self.height))
         img = np.multiply(img, 1 / 255.0)  # [height,width,channel]
@@ -135,13 +138,13 @@ def gen_img(self):
     def get_next_batch(self, batch_size=128):
         batch_x = np.zeros([batch_size, self.width * self.height])
         batch_y = np.zeros([batch_size, self.len * self.max_size])
-        labels  = []
+        labels = []
         for i in range(batch_size):
             image, text, vec = self.gen_img()
             labels.append(text)
             batch_x[i, :] = image.reshape((self.width * self.height))
             batch_y[i, :] = vec
-        return batch_x, batch_y,labels
+        return batch_x, batch_y, labels
 
     def get_predict_source(self):
         items = os.listdir(self.predict_path)
@@ -150,69 +153,44 @@ def get_predict_source(self):
         textes = []
         for i, item in enumerate(items):
             textes.append(item.split("/")[-1])
-            img = cv2.imread(os.path.join(path,item), cv2.IMREAD_GRAYSCALE)
-            img = cv2.resize(img,(self.width,self.height))
+            img = cv2.imread(os.path.join(self.predict_path, item), cv2.IMREAD_GRAYSCALE)
+            img = cv2.resize(img, (self.width, self.height))
             batch_x[i, :] = img.reshape((self.width * self.height))
 
-        return batch_x,textes
+        return batch_x, textes
 
-
-    def text2vec(self,text):
+    def text2vec(self, text):
         text_len = len(text)
         if text_len > self.max_size:
-            raise ValueError('字符最长 ',self.max_size,'个字符')
+            raise ValueError('字符最长 ', self.max_size, '个字符')
         vector = np.zeros(self.max_size * self.len)
         for i, c in enumerate(text):
             idx = i * self.len + chars_indexes[c]
             vector[idx] = 1
         return vector
 
     # 向量转回文本
-    def vec2text(self,vec):
+    def vec2text(self, vec):
         char_pos = vec.nonzero()[0]
         text = []
         for i, c in enumerate(char_pos):
             char_idx = c % self.len
             text.append(chars[char_idx])
         return "".join(text)
 
-    def genBatch(self, batchSize, outputPath):
-        if (not os.path.exists(outputPath)):
-            os.mkdir(outputPath)
-        for i in range(batchSize):
-            # plateStr = self.genPlateString(-1, -1)
-            plateStr, label = self.get_text()
-            img = self.generate(plateStr)
-            img = cv2.resize(img, (self.width, self.height))
-            filename = os.path.join(outputPath, str(i).zfill(4) + '.' + "".join(plateStr) + ".jpg")
-            cv2.imwrite(filename, img)
-
-    def genBatch_v2(self, batchSize, outputPath):
+    def gen_Batch(self, batchSize, outputPath):
         if (not os.path.exists(outputPath)):
             os.mkdir(outputPath)
 
         for i in range(batchSize):
             # plateStr = self.genPlateString(-1, -1)
-            plateStr, label = self.get_text_v2()
+            plateStr, label = self.get_text()
             img = self.generate(plateStr)
             img = cv2.resize(img, (self.width, self.height))
             filename = os.path.join(outputPath, str(i).zfill(4) + '.' + "".join(plateStr) + ".png")
             cv2.imwrite(filename, img)
 
     def get_text(self):
-        text = []
-        label = []
-        text += (self.get_random_char(STATE_CODE, 1))  # 生成车牌
-        text += (self.get_random_char(chars, 2))
-        text += (self.get_random_char(FOUR_CODE, 1))  # 4
-        text += (self.get_random_char(chars, 4))  # 8
-        text += (self.get_random_char(NINE_CODE, 1))  # 9
-        text += (self.get_random_char(chars, 8))
-        for item in text:
-            label.append(chars_indexes[item])
-        return ''.join(text), np.array(label)
-
-    def get_text_v2(self):
         """
          # 随机生成字串，长度固定
         # 返回text,及对应的向量
@@ -226,11 +204,6 @@ def get_text_v2(self):
         text += (self.get_random_char(chars, 4))  # 8
         text += (self.get_random_char(NINE_CODE, 1))  # 9
         text += (self.get_random_char(chars, 8))
-        # size = random.randint(1, self.max_size)
-        # size = self.max_size
-        # for i, c in enumerate(text):
-        #     vec = self.text2vec(c)
-        #     vecs[i * self.len:(i + 1) * self.len] = np.copy(vec)
         return text, self.text2vec(text)
 
     def get_random_char(self, chars, size):
@@ -240,21 +213,11 @@ def get_random_char(self, chars, size):
         for i in ints: result.append(chars[i])
         return result
 
-# generator = GenVin("font.ttf", 49, 258, fontSize=30,NoPlates=bad_path)
-# if __name__ == '__main__':
-#     generator = GenVin("./font/DejaVuSansMono.ttf",80,440)
-#     print("".join(generator.get_text()))
-#     generator = GenVin("./font/Bitter-Regular.ttf")
-#     generator = GenVin("./font/SimHei.ttf")
-#     label,img = generator.gen_sample()
-#     print(label)d
-#     cv2.imshow("test",img)
-#     cv2.waitKeyEx(0)
-#     generator.genBatch_v2(3, "../../data/train/energy")
-#     path = "/Volumes/doc/projects/ml/ocr_tensorflow_cnn/data/train/energy/"
-#     for item in os.listdir(path):
-#         img = cv2.imread(os.path.join(path,item), cv2.IMREAD_GRAYSCALE)
-#         print(img.shape)
-#         cv2.imshow("123", img)
-#         cv2.waitKey(0)
 
+generator = GenVin(49, 258)
+if __name__ == '__main__':
+    # generator = GenVin("./font/DejaVuSansMono.ttf",80,440)
+    # print("".join(generator.get_text()))
+    # generator = GenVin("./font/Bitter-Regular.ttf")
+    # generator = GenVin("./font/SimHei.ttf")
+    generator.gen_Batch(3, test_gen_path)

readme.md

@@ -1,3 +1,5 @@
 # vin recognize
 > 车辆 vin 码识别
 
+## 关键在于样本的生成
+

src/net_work.py

@@ -6,8 +6,6 @@
 import tensorflow as tf
 from lib.gen.generator import GenVin
 import numpy as np
-from lib.logger import logger
-
 
 class NetWork():