diff --git a/docs/TensorFlow2.x/img/output_23_1.png b/docs/TensorFlow2.x/img/output_23_1.png
new file mode 100644
index 00000000..a9111e05
Binary files /dev/null and b/docs/TensorFlow2.x/img/output_23_1.png differ
diff --git a/docs/TensorFlow2.x/img/output_45_0.png b/docs/TensorFlow2.x/img/output_45_0.png
new file mode 100644
index 00000000..a2d05f48
Binary files /dev/null and b/docs/TensorFlow2.x/img/output_45_0.png differ
diff --git a/docs/TensorFlow2.x/img/output_45_1.png b/docs/TensorFlow2.x/img/output_45_1.png
new file mode 100644
index 00000000..bcccd8f2
Binary files /dev/null and b/docs/TensorFlow2.x/img/output_45_1.png differ
diff --git a/docs/TensorFlow2.x/img/output_47_1.png b/docs/TensorFlow2.x/img/output_47_1.png
new file mode 100644
index 00000000..8de91ff5
Binary files /dev/null and b/docs/TensorFlow2.x/img/output_47_1.png differ
diff --git a/docs/TensorFlow2.x/img/output_47_2.png b/docs/TensorFlow2.x/img/output_47_2.png
new file mode 100644
index 00000000..0a10a33f
Binary files /dev/null and b/docs/TensorFlow2.x/img/output_47_2.png differ
diff --git a/docs/TensorFlow2.x/img/output_51_1.png b/docs/TensorFlow2.x/img/output_51_1.png
new file mode 100644
index 00000000..293c0449
Binary files /dev/null and b/docs/TensorFlow2.x/img/output_51_1.png differ
diff --git a/docs/TensorFlow2.x/img/output_53_0.png b/docs/TensorFlow2.x/img/output_53_0.png
new file mode 100644
index 00000000..00ee5764
Binary files /dev/null and b/docs/TensorFlow2.x/img/output_53_0.png differ
diff --git a/docs/TensorFlow2.x/img/overfit_and_underfit__1.png b/docs/TensorFlow2.x/img/overfit_and_underfit__1.png
new file mode 100644
index 00000000..f3a7079d
Binary files /dev/null and b/docs/TensorFlow2.x/img/overfit_and_underfit__1.png differ
diff --git a/docs/TensorFlow2.x/img/overfit_and_underfit__2.png b/docs/TensorFlow2.x/img/overfit_and_underfit__2.png
new file mode 100644
index 00000000..f6d4ff7b
Binary files /dev/null and b/docs/TensorFlow2.x/img/overfit_and_underfit__2.png differ
diff --git a/docs/TensorFlow2.x/img/overfit_and_underfit__3.png b/docs/TensorFlow2.x/img/overfit_and_underfit__3.png
new file mode 100644
index 00000000..a4a22dbf
Binary files /dev/null and b/docs/TensorFlow2.x/img/overfit_and_underfit__3.png differ
diff --git a/docs/TensorFlow2.x/img/overfit_and_underfit__4.png b/docs/TensorFlow2.x/img/overfit_and_underfit__4.png
new file mode 100644
index 00000000..8f0acf88
Binary files /dev/null and b/docs/TensorFlow2.x/img/overfit_and_underfit__4.png differ
diff --git a/docs/TensorFlow2.x/实战项目_2_汽车燃油效率.md b/docs/TensorFlow2.x/实战项目_2_汽车燃油效率.md
index 9f1a6459..84184d38 100644
--- a/docs/TensorFlow2.x/实战项目_2_汽车燃油效率.md
+++ b/docs/TensorFlow2.x/实战项目_2_汽车燃油效率.md
@@ -333,7 +333,7 @@ sns.pairplot(train_dataset[["MPG", "Cylinders", "Displacement", "Weight"]], diag
 
 
 
-![png](http://data.apachecn.org/img/AiLearning/TensorFlow2.x/output_23_1.png)
+![png](img/output_23_1.png)
 
 
 也可以查看总体的数据统计:
@@ -728,11 +728,11 @@ plot_history(history)
 ```
 
 
-![png](http://data.apachecn.org/img/AiLearning/TensorFlow2.x/output_45_0.png)
+![png](img/output_45_0.png)
 
 
 
-![png](http://data.apachecn.org/img/AiLearning/TensorFlow2.x/output_45_1.png)
+![png](img/output_45_1.png)
 
 
 该图表显示在约100个 epochs 之后误差非但没有改进，反而出现恶化。 让我们更新 `model.fit` 调用，当验证值没有提高上是自动停止训练。
@@ -758,11 +758,11 @@ plot_history(history)
     ........................................................
 
 
-![png](http://data.apachecn.org/img/AiLearning/TensorFlow2.x/output_47_1.png)
+![png](img/output_47_1.png)
 
 
 
-![png](http://data.apachecn.org/img/AiLearning/TensorFlow2.x/output_47_2.png)
+![png](img/output_47_2.png)
 
 
 如图所示，验证集中的平均的误差通常在 +/- 2 MPG左右。 这个结果好么？ 我们将决定权留给你。
@@ -804,7 +804,7 @@ _ = plt.plot([-100, 100], [-100, 100])
 
 
 
-![png](http://data.apachecn.org/img/AiLearning/TensorFlow2.x/output_51_1.png)
+![png](img/output_51_1.png)
 
 
 这看起来我们的模型预测得相当好。我们来看下误差分布。
@@ -818,7 +818,7 @@ _ = plt.ylabel("Count")
 ```
 
 
-![png](http://data.apachecn.org/img/AiLearning/TensorFlow2.x/output_53_0.png)
+![png](img/output_53_0.png)
 
 
 它不是完全的高斯分布，但我们可以推断出，这是因为样本的数量很小所导致的。
diff --git a/docs/TensorFlow2.x/实战项目_3_优化_过拟合和欠拟合.md b/docs/TensorFlow2.x/实战项目_3_优化_过拟合和欠拟合.md
index a5a4abcb..4039ae2a 100644
--- a/docs/TensorFlow2.x/实战项目_3_优化_过拟合和欠拟合.md
+++ b/docs/TensorFlow2.x/实战项目_3_优化_过拟合和欠拟合.md
@@ -61,7 +61,7 @@ test_data = multi_hot_sequences(test_data, dimension=NUM_WORDS)
 plt.plot(train_data[0])
 ```
 
-![png](http://data.apachecn.org/img/AiLearning/TensorFlow2.x/overfit_and_underfit__1.png)
+![png](img/overfit_and_underfit__1.png)
 
 ## 证明过拟合
 
@@ -327,7 +327,7 @@ plot_history([('baseline', baseline_history),
               ('bigger', bigger_history)])
 ```
 
-![png](http://data.apachecn.org/img/AiLearning/TensorFlow2.x/overfit_and_underfit__2.png)
+![png](img/overfit_and_underfit__2.png)
 
 请注意，较大的网络仅在一个时期后就开始过拟合，而且过拟合严重。网络的容量越多，将能够更快地对训练数据进行建模（导致较低的训练损失），但网络越容易过拟合（导致训练和验证损失之间存在较大差异）。
 
@@ -421,7 +421,7 @@ plot_history([('baseline', baseline_history),
               ('l2', l2_model_history)])
 ```
 
-![png](http://data.apachecn.org/img/AiLearning/TensorFlow2.x/overfit_and_underfit__3.png)
+![png](img/overfit_and_underfit__3.png)
 
 如您所见，即使两个模型具有相同数量的参数，L2正则化模型也比基线模型具有更高的抗过度拟合能力。
 
@@ -504,7 +504,7 @@ plot_history([('baseline', baseline_history),
               ('dropout', dpt_model_history)])
 ```
 
-![png](http://data.apachecn.org/img/AiLearning/TensorFlow2.x/overfit_and_underfit__4.png)
+![png](img/overfit_and_underfit__4.png)
 
 添加 dropout 是对基线模型的明显改进。