基于Tensorflow的人工智能Mnist图像识别训练实战

基于Tensorflow的人工智能Mnist图像识别训练实战记录备忘

工作目录： ~/worker/ai/MNIST

操作系统：Ubuntu16.04，也可以移植到windows上的Anaconda python环境中去

user@ubuntu1:~/worker/ai$ tree
.
├── MNIST
│   ├── input_data.py
│   ├── input_data.pyc
│   ├── MNIST_data
│   │   ├── t10k-images-idx3-ubyte.gz
│   │   ├── t10k-labels-idx1-ubyte.gz
│   │   ├── train-images-idx3-ubyte.gz
│   │   └── train-labels-idx1-ubyte.gz
│   └── study.py
└── MNIST_data.zip

确认已经安装了python，查看python的版本如下：

user@ubuntu1: ~$ python -V
Python 2.7.12

安装tensorflow：pip install tensorflow

命令pip list 查看确认已经安装了tensorflow

......
tensorboard                   1.13.1
tensorflow                    1.13.1
tensorflow-estimator          1.13.0
termcolor                     1.1.0
tushare                       1.2.21
You are using pip version 19.0.3, however version 19.2 is available.
You should consider upgrading via the 'pip install --upgrade pip' command.

最后执行python study.py，程序先完成9900次神经网络训练，然后进行一次图像识别测试 ，识别准确率达到0.9907

......
Instructions for updating:
Use tf.cast instead.
step 0,train_accuracy= 0.1
step 100,train_accuracy= 0.78
step 200,train_accuracy= 0.94
step 300,train_accuracy= 0.88
step 400,train_accuracy= 0.94
step 500,train_accuracy= 0.88
step 600,train_accuracy= 0.98
step 700,train_accuracy= 0.94
step 800,train_accuracy= 0.86
step 900,train_accuracy= 0.98
step 1000,train_accuracy= 1
......

step 9200,train_accuracy= 1
step 9300,train_accuracy= 1
step 9400,train_accuracy= 1
step 9500,train_accuracy= 1
step 9600,train_accuracy= 1
step 9700,train_accuracy= 0.96
step 9800,train_accuracy= 0.98
step 9900,train_accuracy= 1
2019-07-23 22:27:46.372737: W tensorflow/core/framework/allocator.cc:124] Allocation of 1003520000 exceeds 10% of system memory.
test_accuracy= 0.9907