diff --git a/README.md b/README.md
index c8fe343902cc321a71b6fc1d73206969d2e95034..8ecde528831312d11d5c4c0988c0682c9f65d386 100644
--- a/README.md
+++ b/README.md
@@ -40,7 +40,15 @@ R. Roscher, M. Volpi, C. Mallet, L. Drees, and J. D. Wegner, “Semcity toulouse
 
 The processing workflow to build the classification problem is in this Jupyter Notebook 👉 Build_train_and_test_sets.ipynb
 
-The specific training and test sets that we used for the experiments in the paper are in the folder 👉 experiments 
+The specific data that we used for the experiments in the paper are in the folder 👉 experiments/data
+
+Training set:
+- X_train_tile_4_tiny.npy
+- Y_train_tile_4_tiny.npy
+
+Test set:
+- X_test_tile_8_subregion.npy
+- Y_test_tile_8_subregion.npy
 
 📐 Now you can proceed in two was: