AAM ProjectOut Fitting

Active Appearance Models (AAM) Fitting using the Project Out algorithm.

Category:

Science & Technology

Tags:

• AAM

[pedroadmartins]

hi dailydols, Yes the SIC algorithm works better when the light conditions change. The SIC optimizes both shape and appearance parameters at once. Other algorithms like the Project Out (PO) our the Normalization Inverse Composition (NIC) only search for the shape parameters. Note: SIC is much slower than the PO or the NIC.

[pedroadmartins]

hi dailydols, The video uses ProjectOut plus the inverse compositional update. Both proposed by I.Matthews and S.Baker. (AAM Revisited).

Note that the ProjectOut is only a way to remove the appearance components from the SD images. The Inverse Composition is the way the (shape) parameters are updated.

[dailydols]

I`m still working on better tolerance to light changes. Is  SIC much better on that?

[dailydols]

are you still using the PO approach or did you change to "inverse composition" (IC).

there is a paper from "baker" about it, it says that IC is faster and more stable.

"active appearance models revised"

[ortontim]

Hi, Im trying to do something similar at the moment. could you explain how achieved this? AAM-API? Matlab? openCV?

Also what algorithms are you using?

Thanks - good vid

[ggiorgana]

Hi, nice Job. Did you publish this project? If so, may you please tell me the title of the research paper?

If you have some version that we can download and try out, may you also please tell me the link the of the website.

Thanks in advance!