Frame label interface Demo
In the frame labeling task, we ask annotators to name all the actions in a sequence, and precisely label the start and end of each action, including transitions between distinct actions.
Above, we provide a link to our frame labeling web interface with 2 mocap sequences.
In the first sequence, we provide an already-labeled sample from BABEL. It illustrates a high level of annotation detail.
The second sequence is unlabeled, so viewers can attempt the annotation task.
Sequence label interface Demo
The sequence labeling task collects two types of labels -- (a) If the sequence contains multiple actions (yes/no), and (b) an action label that describes the overall action in the sequence.
Above, we provide a link to our sequence labeling web interface with 4 unlabeled mocap sequences.