微信小程序+中草药分类+爬虫+keras

本文介绍: 数据爬取：使用爬虫爬取百度图片，可以自己定义要爬取的中草药种类、数量等信息。模型训练使用基于keras训练分类模型，模型可以修改，例如：ResNet50系列，MobileNet系列等，支持在gpu、cpu 训练。后台 flask 部署：使用 flask将模型部署到后台，提供ip 地址和端口号前端微信小程序：制作前端的微信小程序页面，将图片传输给后端，并且将分类结果返回到前端展示。

本项目使用深度学习模型，训练5种中药材数据集，然后将其集成到微信小程序，通过微信小程序拍照，将图片传输给后端，后端将返回的结果展示到前端页面，项目主要包含以下内容：

使用requests进行爬虫
示例：

  for i in range(30):
            image_url = result['data'][i]['middleURL']
            image_name = "%d.jpg" % count
            response = requests.get(image_url, headers=headers, stream=True, timeout=10)
            with open(os.path.join(download_path, image_name), 'wb') as f:
                f.write(response.content)
            count += 1

爬取输入参数，可以自己输入爬取哪些中草药，输入到list里面即可，下面展示只爬取两种中草药。

# 设置搜索关键字和爬取图片的数量
name_list = ['枸杞','金银花']
save_path = "data_爬虫"
page_num = 1 #爬取多少页，每页30个
for keyword in name_list:
    get_images(save_path, keyword, page_num)

在这里插入图片描述

此处，我们分别使用keras版本进行训练和验证，具体代码和结果展示如下：

from keras.preprocessing.image import ImageDataGenerator
from keras.optimizers import Adam
from keras.applications import MobileNetV2
from keras.layers import GlobalAveragePooling2D, Dense
from keras.models import Sequential
import json

# 定义ImageDataGenerator
datagen = ImageDataGenerator(
    rescale=1./255,
    shear_range=0.2,
    zoom_range=0.2,
    horizontal_flip=True,
    validation_split=0.2  # 设置验证集的比例
)

base_model = MobileNetV2(weights='imagenet', include_top=False, input_shape=(224, 224, 3))

model = Sequential([
    base_model,
    GlobalAveragePooling2D(),
    Dense(128, activation='relu'),
    Dense(num_classes, activation='softmax')
])

# 训练模型
model.fit(
    train_generator,
    steps_per_epoch=train_generator.samples // batch_size,
    epochs=10,
    validation_data=validation_generator,
    validation_steps=validation_generator.samples // batch_size)

# 步骤6: 评估模型性能
eval_result = model.evaluate(validation_generator)
print(f"Test accuracy: {eval_result[1]*100:.2f}%")

36/36 [==============================] - 22s 449ms/step - loss: 0.7144 - accuracy: 0.7664 - val_loss: 0.7706 - val_accuracy: 0.7278
Epoch 2/10
36/36 [==============================] - 13s 352ms/step - loss: 0.1504 - accuracy: 0.9601 - val_loss: 0.5325 - val_accuracy: 0.8278
Epoch 3/10
36/36 [==============================] - 13s 352ms/step - loss: 0.0959 - accuracy: 0.9829 - val_loss: 0.2743 - val_accuracy: 0.9222
Epoch 4/10
36/36 [==============================] - 13s 351ms/step - loss: 0.0896 - accuracy: 0.9758 - val_loss: 0.3960 - val_accuracy: 0.8500
Epoch 5/10
36/36 [==============================] - 13s 354ms/step - loss: 0.0743 - accuracy: 0.9758 - val_loss: 0.2853 - val_accuracy: 0.9111
Epoch 6/10
36/36 [==============================] - 13s 351ms/step - loss: 0.0525 - accuracy: 0.9829 - val_loss: 0.2473 - val_accuracy: 0.9222

import cv2
import numpy as np
import json
from keras.models import load_model

def get_img(img_path,img_width, img_height ):
    img = cv2.imread(img_path)
    img = cv2.resize(img, (img_width, img_height))  # 调整图像大小
    img = img.astype("float") / 255.0  # 数据预处理，确保与训练时一致
    img = np.expand_dims(img, axis=0)
    return img
    
img_width = 224
img_height = 224
model = load_model(r'E:project1-zhongcaoyaomodel-keras.h5')
print(class_indict)
img_file_path = 'data_all/baihe/b (20).jpg'
classify_img = get_img(img_file_path,img_width, img_height)
results = np.squeeze(model.predict(classify_img)).astype(np.float64)  # 获得预测结果(注意:1.降维2.json中的小数类型为float)
predict_class = np.argmax(results)  # 获得预测结果中置信度最大值所对应的下标

DecodeError: 'utf-8' codec can't decode byte 0xc6 in position 10: invalid continuation byte

app = flask.Flask(__name__)
idx2class = {0:"百合",1:"党参",2:"枸杞",3:"槐花",4:"金银花"}

idx2info ={}
# 导入药效信息
with open("info.txt", "r", encoding="UTF-8") as fin:
    lines = fin.readlines()
    for line in lines:
        idx = int(line.strip().split(":")[0])
        info = line.strip().split(":")[1]
        idx2info[idx] = info

img_bytes = flask.request.form.get('picture') # 获取值
image = base64.b64decode(img_bytes)# 编码转换
image = Image.open(io.BytesIO(image))
classify_img = prepare_image(image,224,224) # 预处理图像
results = np.squeeze(model.predict(classify_img)).astype(np.float64)  # 获得预测结果(注意:1.降维2.json中的小数类型为float)
predicted_idx = np.argmax(results)  # 获得预测结果中置信度最大值所对应的下标
score = results[predicted_idx]
label_name = idx2class[predicted_idx]
label_info = idx2info[predicted_idx]

wx.request({
          url: 'http://127.0.0.1:8080/predict', //本地服务器地址
    
          method: 'POST',

          header: {
            'content-type': 'application/x-www-form-urlencoded'
          },
    
          data: {
            "picture": that.data.picture,
          },
          
          success: (res)=>{
            that.setData({
                class_name: res.data['class_name'],
                prob: res.data['prob'],
                info:res.data['info']
            })