diff --git a/README.md b/README.md
index b124214..4248c5c 100644
--- a/README.md
+++ b/README.md
@@ -14,6 +14,7 @@ See the [project page](https://lmb.informatik.uni-freiburg.de/projects/freihand/
 
 # New:exclamation:
 An extendend version of this dataset, with calibration data and multiple-views, is released in [HanCo](https://lmb.informatik.uni-freiburg.de/projects/contra-hand/).
+Due to ungoing problems with the Codalab evaluation server we have decided to release the evaluation split annotations publicly on our [dataset page](https://lmb.informatik.uni-freiburg.de/resources/datasets/FreihandDataset.en.html).
 
 
 # Basic setup
@@ -68,6 +69,8 @@ The script provides a couple of other parameters you might want to try. Note tha
     
 # Evaluate on the dataset
 
+**Update**: Due to ungoing problems with the Codalab evaluation server we have decided to release the evaluation split annotations publicly on our [dataset page](https://lmb.informatik.uni-freiburg.de/resources/datasets/FreihandDataset.en.html).
+
 In order to ensure a fair and consistent protocol, evaluation of your algorithm on FreiHAND is handled through Codalab.
  
 1. Make predictions for the evaluation dataset. The code provided here predicts zeros for all joints and vertices.