System architecture for indexing regions in keyframes
Abstract
This paper describes the design of an indexing system for a video database. The system uses region-based manual an- notations of keyframes to create models to automatically annotate new keyframes also at the region level. The pre- sented architecture includes user interfaces for training and querying the system, internal databases to manage ingested content and modelled semantic classes, as well as communi- cation interfaces to allow the system interconnection. The scheme is designed to work as a plug-in to an external Mul- timedia Asset Management (MAM) system.
Full text
SYSTEM ARCHITECTURE FOR INDEXING
REGIONS IN KEYFRAMES
Xa ie Gi ó and Fe an Ma qués
Goal
P oposed solu ion
Key ames
Videos
P esen e
Head Body
Hai Face
Funded by:
● Design an a chi ec u e capable o au oma ically index a he key ame egion-le el he
ideo asse o a con en p oduce .
● Plug-in o an exis ing Mul imedia Asse Managemen (MAM)
Mul imedia Asse Manage (MAM)
Pa i ion
BPT
Visual
desc ip o s
Head Body
Rela ions
Seman ic
Pe cep ual
+
+
+
+-
--
-
Au oma ic
Manual
Con en DB Knowledge DB Models DB Ins ances DB
Index
Keywo d Video ID
P esen e ID1, ID3, ID5
C owd ID2
A s ID4
Tes
T ain
Objec pa s
Que y by example
G aphical Use In e aces (GUIs)
Manual anno a ion
Au oma ic Indexe plug-in
Que y by keywo dCon en inges