Better documentation on the dump

flandmark/script/annotate.py

@@ -4,6 +4,44 @@
 # Fri 21 Sep 2012 10:43:12 CEST 
 
 """Annotates videos, dumps annotations as text files.
+
+The text files will contain one line per annotated frame. Each line contains a
+single detected face and associated landmarks. Only the biggest detection found
+on the frame is considered. The entries in each line are:
+
+[0]
+  The frame number (starting from 0)
+
+[1:5] 4 items
+  The bounding-box coordinates as detected by OpenCV (x, y, width, height)
+
+[5:] 8 pairs
+  Each pair corresponds to a keypoint in the order defined by the model:
+
+  [5:7]
+    Face center (as defined by the OpenCV detected bounding box)
+
+  [7:9]
+    Canthus-rl (inner corner of the right eye). Note: The "right eye" means
+    the right eye at face w.r.t. itself - that is the left eye in the image.
+
+  [9:11]
+    Canthus-lr (inner corder of the left eye)
+
+  [11:13]
+    Mouth-corner-r (right corner of the mouth)
+
+  [13:15]
+    Mouth-corner-l (left corner of the mouth)
+
+  [15:17]
+    Canthus-rr (outer corner of the right eye)
+
+  [17:19]
+    Canthus-ll (outer corner of the left eye)
+
+  [19:21]
+    Nose
 """
 
 import os