: In a technical sense, the "features" for this video would be the optical flow (movement between frames) or deep learning descriptors (like I3D or C3D features) extracted from the pixels to help a machine "see" the action. Common Contexts
: The "g" often stands for a specific "group" or category of action. In the UCF101 dataset, for instance, a file like v_ApplyEyeMakeup_g05_c04.avi (often converted to .mp4 ) would represent the activity Apply Eye Makeup . g_054.mp4
: Activities like Baby Crawling , Bench Press , or Biking . Kinetics : A much larger set of human-object interactions. : In a technical sense, the "features" for
If you are working with a specific project, "g_054.mp4" likely belongs to one of these actions: : Activities like Baby Crawling , Bench Press , or Biking
In these academic contexts, "features" usually refer to the following:
: The g_05 or g_054 portion indicates the group ID . Researchers use these groups to ensure that videos of the same person or setting are kept together in either the training or testing set to prevent "data leakage."