[Work Log] Implementing single-view tree fitting
January 26, 2015
Implemented generic kernel function
Implemented and tested derivative of generic kernel function
Current: step through blind_tree_fitting.m
MAP point estimation is failing. Is this because we're using the wrong prior?
Next: Train new prior
Then: resume stepping through blind_tree_fitting.m
Current: train new prior
done. implemented with analytical gradient; converges in 25 iterations.
Saved to 'training_output/single_view_prior.mat'
Current: step through blind_tree_fitting.m
MAP point estimation is failing. Is this because we're using the wrong prior?
We had omitted the noise variance. Also, matrix transpose error.
TODO: incorporate mean.
Later:
Test derivative estimation; normal direction estimation
plot each curve, and tangent/normal