Upload 2 files

README.md

@@ -1,3 +1,18 @@
----
-license: cc-by-4.0
----
+# Pouring Demonstrations Dataset, v1.0.
+Pierre Sermanet, sermanet@google.com
+
+## License
+This data is licensed by Google Inc. under a Creative Commons Attribution 4.0 International License.
+
+## Structure
+
+A train and test sets are provided with the following structure:
+
+videos/: original .mov videos
+frames/: extracted frames from videos/
+
+## Labels
+
+Temporal segmentation groundtruth is provided in segmentation_groundtruth.py for 2 and 3 segments
+for each video. Each segment corresponds to one semantic step for the demonstration.
+

segmentation_groundtruth.py

@@ -0,0 +1,28 @@
+# Temporal segmentation groundtruth for pouring task sequences.
+# '2' indicates 2 segments, e.g. step-1 corresponds to frames [0, 32] and
+# step-2 to frames [33,last_frame].
+# '3' indicates 3 segments segmentation.
+segmentation_groundtruth = {
+   # This does not include the 'miss' videos for which segmentation cannot be defined.
+  'test': {
+    'pouring_001': {'2': [32], '3': [8, 40]},
+    'pouring_004': {'2': [27], '3': [11, 32]},
+    'pouring_007': {'2': [23], '3': [11, 26]},
+    'pouring_009': {'2': [32], '3': [9, 39]},
+    'pouring_012': {'2': [20], '3': [7, 23]},
+    'pouring_016': {'2': [24], '3': [12, 28]},
+  },
+  'train': {
+    'pouring_002': {'2': [25], '3': [11, 35]},
+    'pouring_003': {'2': [32], '3': [12, 38]},
+    'pouring_005': {'2': [27], '3': [10, 32]},
+    'pouring_006': {'2': [21], '3': [10, 27]},
+    'pouring_008': {'2': [18], '3': [10, 23]},
+    'pouring_010': {'2': [22], '3': [9, 30]},
+    'pouring_011': {'2': [20], '3': [10, 27]},
+    'pouring_013': {'2': [22], '3': [10, 29]},
+    'pouring_014': {'2': [20], '3': [8, 25]},
+    'pouring_015': {'2': [25], '3': [9, 35]},
+    'pouring_017': {'2': [18], '3': [9, 20]},
+  },
+}