diff --git a/README.md b/README.md
index 704081dc19a8db1ce2552f7066ecfad239cf3c3e..b359e66dfdb8baf7a90cefdc627aefd7271715a7 100644
--- a/README.md
+++ b/README.md
@@ -11,6 +11,7 @@ To access the CIFAR-10 database and obtain the necessary data, you can visit htt
 **KNN**
 
 The result obtained from the k-nearest neighbors (KNN) is presented as the accuracy, showcasing how it varies with changes in the value of k, ranging from 1 to 20.
+Here we can see that the graph decrease, with a mean accuracy of 35% which is not very good,but expected, because images are not always the same ( not only cats or cars but all mixed)
 
 ![training accuracy knn](results/KNN.png)
 
@@ -18,6 +19,7 @@ The result obtained from the k-nearest neighbors (KNN) is presented as the accur
 **NN**
 
 The output from the neural networks (NN) illustrates the progression of learning accuracy over different learning epochs.
+Here we can see that the graph increase, because all along the epochs, the system is training with the same type of image, at the contrary to the first case.
 
 ![Training accuracy mlp](results/mlp_training_accuracy.png)