en-de
en-es
en-fr
en-sl
en
en-zh
0.25
0.5
0.75
1.25
1.5
1.75
2
Exploiting Problem Structure for Efficient Discrete Optimization
Published on Jan 25, 20125572 Views
Many problems in computer vision and machine learning require inferring the most probable states of certain hidden or unobserved variables. This inference problem can be formulated in terms of minimiz
Related categories
Chapter list
Exploiting Problem Structure for Efficient Discrete Optimization00:00
Image Segmentation - 101:29
Image Segmentation - 203:14
Image Segmentation - 304:23
Image Segmentation - 404:40
Image Segmentation - 504:43
Image Segmentation - 605:39
Energy Minimization Problems06:01
Submodular Functions: Definition - 108:10
Submodular Functions: Definition - 209:44
Submodular Functions: Definition - 311:23
Submodular Functions12:44
Minimizing Submodular Functions13:17
So how does this work? 14:04
Flow and Reparametrization - 114:28
Flow and Reparametrization - 215:34
Flow and Reparametrization - 316:59
Flow and Reparametrization - 417:08
Flow and Reparametrization - 517:15
Flow and Reparametrization - 617:16
Flow and Reparametrization - 717:19
Flow and Reparametrization - 817:29
Flow and Reparametrization - 918:14
Flow and Reparametrization - 1018:20
Flow and Reparametrization - 1119:20
Flow and Reparametrization - 1220:10
Demo20:18
.. So what are the challenges? 21:23
Need Richer Models21:27
3,600,000,000 Pixels21:58
Speed and Scalability - 122:37
Speed and Scalability - 222:56
Part I : Exploiting Problem Structure for Efficiency23:29
Image Segmentation in Videos - 123:41
Image Segmentation in Videos - 223:46
Image Segmentation in Videos - 324:08
Dynamic Energy Minimization - 124:31
Dynamic Energy Minimization - 225:51
Hybrid Algorithms - 126:44
Hybrid Algorithms - 227:06
Hybrid Algorithms - 327:07
Hybrid Algorithms - 427:57
Making Inference algorithms adapt to the problem28:51
General Labeling Problem - 129:06
General Labeling Problem - 230:09
General Labeling Problem - 330:16
MAP Inference as an IP30:26
LP Relaxation of MAP Inference31:02
Solving the LP Relaxations - 132:07
Solving the LP Relaxations - 233:43
Solving the LP Relaxations - 334:00
Solving the LP Relaxations - 434:14
Can we make inference algorithms adapt to the `current’ problem?34:58
Local Primal-Dual Gaps35:06
Focused Inference - 137:44
Energy-Aware Message-Passing38:14
Message Passing38:15
Intelligent Message Scheduling - 138:44
Intelligent Message Scheduling - 239:24
Intelligent Message Scheduling - 339:46
Focused Inference - 240:02
Label re-ordering in alpha-Expansion40:03
Graph Cut based Move Making Algorithms - 140:06
Graph Cut based Move Making Algorithms - 240:59
Dynamic Re-ordering of Labels - 142:03
Dynamic Re-ordering of Labels - 242:05
Dynamic Re-ordering of Labels - 342:06
Dynamic Re-ordering of Labels - 442:28
Label re-ordering in alpha-Expansion: Experiments42:43
Focused Inference - 342:47
Tightening LP Relaxations42:48
Hierarchy of LPs42:51
Tightening Relaxation 43:07
Tightening LP Relaxations: Experiments43:34
Part II : Exploiting Problem Structure for Handling Richer Models43:55
Image Segmentation - 144:07
Pn Potts Model for label consistency44:24
Image Segmentation - 244:52
Image Segmentation - 344:53
Higher Order Potentials for Object Segmentation - 144:55
Higher Order Potentials for Object Segmentation - 245:25
Higher Order Potentials for Object Segmentation - 346:01
Results46:01
More Results46:03
Overcoming short-boundary bias - 146:03
Overcoming short-boundary bias - 246:04
Overcoming short-boundary bias - 346:05
Overcoming short-boundary bias - 446:06
Minimizing Higher Order Energy Functions46:06
Higher order to Quadratic - 146:30
Higher order to Quadratic - 246:38
Challenges and Opportunities47:20