diff --git a/alignment/coordinateReg/README.md b/alignment/coordinateReg/README.md
new file mode 100644
index 0000000..fba8197
--- /dev/null
+++ b/alignment/coordinateReg/README.md
@@ -0,0 +1,12 @@
+Two 2D-106 pretrained models are provided to predict landmarks by given loose crop face images.
+
+Model ``2d106det``: Given face detection bounding box, predict landmarks.
+
+Model ``2d106track``: Given landmarks bounding box, predict landmarks.
+
+Model ``2d106det`` is usually used for static image inference, while model ``2d106track`` can be used for video landmarks tracking.
+
+Visualization: TODO
+
+
+Download:  TODO