1 / 7

Challenges in Learning and Inference for MRF and CRF Models

Explore the difficulties faced in learning and inference with Markov Random Fields (MRFs) and Conditional Random Fields (CRFs) models, including issues like limited support, global label inference, and complex graph structures limiting generality. Learn about approximate solutions and the complexity constraints that restrict model complexity.

castroj
Download Presentation

Challenges in Learning and Inference for MRF and CRF Models

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. MRFs (X1,X2) 1 2 3 X3 X1 X2 4 (X2,X3,X3) X4

  2. CRFs Image I (X1,X2,I) 1 2 3 X3 X1 X2 4 (X2,X3,X3,I) X4

  3. CRF MRF Examples X = Image patches [Quattoni et al.] X = Patches on a regular lattice [Kumar]

  4. Examples X = pixels, regions, image [He et al.]

  5. Issues • Inference • Easy only when the planets are aligned • Approximate solutions only  How good are they? • Learning • Difficult and slow • Limits the complexity of the models

  6. Issues • Generality • Can use arbitrary models but limited to restricted models in practice because of inference and learning challenges • Global label inference • Inference over global labeling of the data in theory, but limited propagation across image in practice • Support limited by the complexity of learning and inference • Use of complex graph and clique structure is difficult

More Related