>- **🍨 本文为[🔗365天深度学习训练营](https://mp.weixin.qq.com/s/rbOOmire8OocQ90QM78DRA) 中的学习记录博客**
>- **🍖 原作者:[K同学啊 | 接辅导、项目定制](https://mtyjkh.blog.csdn.net/)**
我的环境:
- 系统环境:WSL2+Ubuntu22.04
- 语言环境:Python3.8.18
- 编译器:vscode+jupyter notebook
- 深度学习环境:TensorFlow2.10.0
- GPU: 5700xt
T4:猴痘病识别
🍺要求:
- 训练过程中保存效果最好的模型参数。(完成)
- 加载最佳模型参数识别本地的一张图片。(完成)
- 调整网络结构使测试集accuracy到达88%(重点)。(最终达到90.65%,完成)
🍻拔高(可选):
- 调整模型参数并观察测试集的准确率变化。(完成)
- 尝试设置动态学习率。(完成)
- 测试集accuracy到达90%。(最终达到90.65%,完成)
首先调周期至100,观察输出:
Epoch 100: val_accuracy did not improve from 0.89486
54/54 [==============================] - 1s 27ms/step - loss: 0.0126 - accuracy: 0.9971 - val_loss: 0.6966 - val_accuracy: 0.8879
测试集准确率0.8879,直接满足基础要求了。。。
然后解决过拟合的问题:
观察样本总数:2142,一开始想用数据增强的手法,结果打开文件夹看了下一张图变了十三次,已经增强过了。。。
那就减少模型复杂度,为模型增加第三个dropout层:
model = models.Sequential([
layers.experimental.preprocessing.Rescaling(1./255, input_shape=(img_height, img_width, 3)),
layers.Conv2D(16, (3, 3), activation='relu', input_shape=(img_height, img_width, 3)), # 卷积层1,卷积核3*3
layers.AveragePooling2D((2, 2)), # 池化层1,2*2采样
layers.Conv2D(32, (3, 3), activation='relu'), # 卷积层2,卷积核3*3
layers.AveragePooling2D((2, 2)), # 池化层2,2*2采样
layers.Dropout(0.3),
layers.Conv2D(64, (3, 3), activation='relu'), # 卷积层3,卷积核3*3
layers.Dropout(0.3),
layers.Flatten(), # Flatten层,连接卷积层与全连接层
layers.Dropout(0.3),
layers.Dense(128, activation='relu'), # 全连接层,特征进一步提取
layers.Dense(num_classes) # 输出层,输出预期结果
])
运行得:
Epoch 100: val_accuracy did not improve from 0.90888
54/54 [==============================] - 2s 28ms/step - loss: 0.0028 - accuracy: 1.0000 - val_loss: 0.5561 - val_accuracy: 0.9065
Output is truncated. View as a scrollable element or open in a text editor. Adjust cell output settings...
验证集准确率0.9065,直接就满足了。
这期有点简单
读取模型识别图片:
直接运行会报错:
{
"name": "ValueError",
"message": "Attempt to convert a value (<PIL.JpegImagePlugin.JpegImageFile image mode=RGB size=224x224 at 0x7FC344568160>) with an unsupported type (<class 'PIL.JpegImagePlugin.JpegImageFile'>) to a Tensor.",
"stack": "---------------------------------------------------------------------------
ValueError Traceback (most recent call last)
/home/wjh/CNN/练习/T4猴痘病识别/T4猴痘病识别.ipynb 单元格 17 line 6
<a href='vscode-notebook-cell://wsl%2Bnewubuntu-22.04/home/wjh/CNN/%E7%BB%83%E4%B9%A0/T4%E7%8C%B4%E7%97%98%E7%97%85%E8%AF%86%E5%88%AB/T4%E7%8C%B4%E7%97%98%E7%97%85%E8%AF%86%E5%88%AB.ipynb#X22sdnNjb2RlLXJlbW90ZQ%3D%3D?line=3'>4</a> # img = Image.open(\"./45-data/Monkeypox/M06_01_04.jpg\") #这里选择你需要预测的图片
<a href='vscode-notebook-cell://wsl%2Bnewubuntu-22.04/home/wjh/CNN/%E7%BB%83%E4%B9%A0/T4%E7%8C%B4%E7%97%98%E7%97%85%E8%AF%86%E5%88%AB/T4%E7%8C%B4%E7%97%98%E7%97%85%E8%AF%86%E5%88%AB.ipynb#X22sdnNjb2RlLXJlbW90ZQ%3D%3D?line=4'>5</a> img = Image.open(\"/home/wjh/CNN/练习/T4猴痘病识别/data/Others/NM15_02_11.jpg\") #这里选择你需要预测的图片
----> <a href='vscode-notebook-cell://wsl%2Bnewubuntu-22.04/home/wjh/CNN/%E7%BB%83%E4%B9%A0/T4%E7%8C%B4%E7%97%98%E7%97%85%E8%AF%86%E5%88%AB/T4%E7%8C%B4%E7%97%98%E7%97%85%E8%AF%86%E5%88%AB.ipynb#X22sdnNjb2RlLXJlbW90ZQ%3D%3D?line=5'>6</a> image = tf.image.resize(img, [img_height, img_width])
<a href='vscode-notebook-cell://wsl%2Bnewubuntu-22.04/home/wjh/CNN/%E7%BB%83%E4%B9%A0/T4%E7%8C%B4%E7%97%98%E7%97%85%E8%AF%86%E5%88%AB/T4%E7%8C%B4%E7%97%98%E7%97%85%E8%AF%86%E5%88%AB.ipynb#X22sdnNjb2RlLXJlbW90ZQ%3D%3D?line=7'>8</a> img_array = tf.expand_dims(image, 0)
<a href='vscode-notebook-cell://wsl%2Bnewubuntu-22.04/home/wjh/CNN/%E7%BB%83%E4%B9%A0/T4%E7%8C%B4%E7%97%98%E7%97%85%E8%AF%86%E5%88%AB/T4%E7%8C%B4%E7%97%98%E7%97%85%E8%AF%86%E5%88%AB.ipynb#X22sdnNjb2RlLXJlbW90ZQ%3D%3D?line=9'>10</a> predictions = model.predict(img_array) # 这里选用你已经训练好的模型
File ~/anaconda3/envs/tensorflow/lib/python3.8/site-packages/tensorflow/python/util/traceback_utils.py:153, in filter_traceback.<locals>.error_handler(*args, **kwargs)
151 except Exception as e:
152 filtered_tb = _process_traceback_frames(e.__traceback__)
--> 153 raise e.with_traceback(filtered_tb) from None
154 finally:
155 del filtered_tb
File ~/anaconda3/envs/tensorflow/lib/python3.8/site-packages/tensorflow/python/framework/constant_op.py:102, in convert_to_eager_tensor(value, ctx, dtype)
100 dtype = dtypes.as_dtype(dtype).as_datatype_enum
101 ctx.ensure_initialized()
--> 102 return ops.EagerTensor(value, ctx.device_name, dtype)
ValueError: Attempt to convert a value (<PIL.JpegImagePlugin.JpegImageFile image mode=RGB size=224x224 at 0x7FC344568160>) with an unsupported type (<class 'PIL.JpegImagePlugin.JpegImageFile'>) to a Tensor."
}
这个报错原因是,tensorflow框架只能输入numpy数组,直接读取jpg图像得到的是PIL图像。
于是要将图像转为numpy数组的格式再进行输入:
from PIL import Image
import numpy as np
import tensorflow as tf
# img = Image.open("./45-data/Monkeypox/M06_01_04.jpg") #这里选择你需要预测的图片
img = Image.open("/home/wjh/CNN/练习/T4猴痘病识别/data/Others/NM15_02_11.jpg") #这里选择你需要预测的图片
img = img.resize((img_width, img_height)) # 调整图像大小
img_array = np.array(img) # 转换为 NumPy 数组
img_array = tf.expand_dims(img_array, 0) # 添加批次维度
predictions = model.predict(img_array) # 这里选用你已经训练好的模型
print("预测结果为:", class_names[np.argmax(predictions)])
输出为:
1/1 [==============================] - 0s 64ms/step
预测结果为: Others
2023-12-13 11:50:23.345980: I tensorflow/core/grappler/optimizers/custom_graph_optimizer_registry.cc:114] Plugin optimizer for device_type GPU is enabled.
成功