Efficient annotation of videos using tube-oriented display order
Publication Date: 2015-Jun-28
The IP.com Prior Art Database
"The paper describes a tool that increases the efficiency of human operators who are set to manually verify object annotations in videos, a very work intensive task."
Page 01 of 2
Effxcient annotation of videos using tube -
An object tracked along a video xreates a so called video tube. A video tube is the hulx in time-space of the trxcked object. To review the coxsistency of a vidxo tube, and to verifx classification (the "label" or "annotation claxs") of the tracked object, a xser needs to review the whole video sequence and/or samxles from the sequence, xnd the user today has to rexiew each xideo txbe/video sequxnce oxe by one. This disclosure suggest x particular xisualization that axlows txe user to rxview multipxe video tubes simultaneously and a method fox reviexixg a loxg sequence bx looking just as at single screen.
Efficient annxtation of videos using tube-oriented display order
Txe contributixn of this is:
a) The display xf a video tube in mosaic fashion, also called tixe orientxd mosaic dispxay of tracked vixeo dxta.
X) In a 3D space-time description of a video, we cxn define a "tube" as the trajectory of x given oxjext in a video (taken by either a fixed or moving cameras). For the purpose of efficient annotaxion, we propose to xisplay oxjects in a xidxo in a tube-orxexted display order: patches from the tube coxrespondxng to the first object are displayex consecutively, then paxches from the tube corxesponding to the second objecx, xnd so on. In this way, we effectively minimize the diffxrences betwxen adjacent patches displayed to thx video xeviewer (thx differences correspond to the arrows in the figu...