en
0.25
0.5
0.75
1.25
1.5
1.75
2
Large-Scale Object Classification using Label Relation Graphs
Published on Oct 29, 20149040 Views
In this paper we study how to perform object classification in a principled way that exploits the rich structure of real world labels. We develop a new model that allows encoding of flexible relations
Related categories
Chapter list
Large-Scale Object Recognition using Label Relation Graphs00:00
Object Classification - 100:01
Object Classification - 200:16
Object Classification - 300:22
Object Classification - 400:30
Object labels have rich relations01:26
Goal: A new classification model02:21
Visual Model + Knowledge Graph02:31
Agenda - 102:55
Agenda - 203:09
Hierarchy and Exclusion (HEX) Graph03:13
Examples of HEX graphs03:50
State Space: Legal label configurations - 104:49
State Space: Legal label configurations - 205:15
State Space: Legal label configurations - 305:29
Agenda - 306:02
HEX Classification Model - 106:10
HEX Classification Model - 206:22
HEX Classification Model - 306:45
HEX Classification Model - 407:15
HEX Classification Model - 507:34
Special Case of HEX Model07:45
Learning08:23
Agenda - 408:57
Naïve Exact Inference is Intractable09:04
Observation 1: Exclusions are good09:24
Observation 2: Equivalent graphs - 109:59
Observation 2: Equivalent graphs - 210:37
HEX Graph Inference10:58
Agenda - 511:30
Exp1: Learning with weak labels - 111:34
Exp1: Learning with weak labels - 211:55
Exp1: Learning with weak labels - 312:15
Exp1: Learning with weak labels - 412:34
Exp2: Zero-Shot Recognition using Object-Attribute Knowledge12:50
Related Work13:46
Conclusions14:23