Automatic landmark detection model for planar shape data
Detailedcode.zip is a zipped folder containing the following Matlab programs:
Main programs
ALDfixed.m
- Metropolis-Hastings sampler for fixed number of landmarks model (inferring \theta)
ALDunknown.m
- reversible jump MCMC for unknown number of landmarks model (inferring k and \theta)
FindBestk.m
- repeated parallel MCMC runs of ALDfixed.m for criterion-based approach for selecting number of landmarks -- uses parallel computing functionality in MATLAB
Sub programs
ClosedIntSqDist.m
- evaluate distance between closed curve and its linear reconstruction through landmarks
curve_to_q.m
- convert a two-dimensional curve to its SRVF representation
InnerProd_Q.m
- compute L2 inner product
MoveProb.m
- evaluates move probabilities for RJMCMC
OpenIntSqDist.m
- evaluate distance between open curve and its linear reconstruction through landmarks
q_to_curve.m
- convert a SRVF to its two-dimensional curve representation
ReSampleCurve.m
- assuming a dense grid, re-samples two-dimensional curves (at equal spacings) to a desired number of points
ShiftF.m
- used to shift the starting point of a closed curve (important for posterior processing of labels)
Wrapper to replicate paper examples from "Automatic Detection and Uncertainty Quantification of Landmarks on Elastic Curves" (https://www.tandfonline.com/doi/full/10.1080/01621459.2018.1527224):
ExampleCode.m
Data
mice.mat
(with help file MiceLabels.txt
)
MPEG7closed.mat
(with help file MPEG7closedLabels.txt
)
ToyCurve.mat
(with help file ToyCurveLabels.txt
)
E-mail Justin Strait ([email protected]) with any questions about the code.