Scalable Video Genre Classification and Event Detection - Multimedia Database Retrieval: Technology and Applications - page 255

Table 9.2 Comparison of event detection models emphasizing feature utilization from both low-
level features and middle-level semantic agents
Event
detection

Visual features

Low-level Middle-level

Year

algorithm Nature of Number of multimodal Global- Local- semantic

category Reference data events features based based agents

Patten-
recognition
model

2003
[ 268 ]

Tennis 5 AV M Ye s No Ye s

2005
[ 267 ]

Four field
sports

2 AV S Ye s No No

2005
[ 269 ]

Soccer 1 n/a n/a n/a Ye s

2009
[ 270 ]

Basketball 5 VM Ye s No No

Semantic
event
model

2002
[ 271 ]

Football 3 VST Ye s No No

2002
[ 272 ]

Baseball 2 VT Ye s No No

2003
[ 257 ]

Soccer 3 VS Ye s No Ye s

2001
[ 246 ]

Basketball 1 AV M T Ye s No No

2003
[ 273 ]

Tennis
soccer

16 AV M T Ye s No Ye s

State event
model

2004
[ 276 ]

Soccer 2 VM Ye s No Ye s

2006
[ 278 ]

Soccer 5 AV M Ye s No Ye s

2007
[ 275 ]

Basketball 5 VT Ye s No Ye s

2008
[ 274 ]

Tennis 4 AV S Ye s No No

2008
[ 277 ]

Soccer 3 VM Ye s No Ye s

2008
[ 279 ]

Soccer
basketball

17 VTS Ye s No Ye s

2009
[ 247 ]

Soccer 6 VMTS Ye s No Ye s

In the “Low-level Multimodal Features” column, various features are utilized, including audio (A),
visual (V), text (T), motion feature (M), and video shot detection (S), as well as an “n/a” label in
the case when no low-level feature mentioned in the related works