- work log 200
- reference 30
- todo 3
- strategy 1
- realization: current likelihood permits bad candidates 1
- meeting notes 1
- notes 1

- Work Log
- Work Log
- Work Log
- Work Log
- Work Log
- Work Log
- Work Log
- Work Log
- Work Log
- Work Log
- Work Log
- Work Log
- Work Log
- Work Log
- Implementing single-view tree fitting
- Work Log
- Work Log
- Work Log - fitting progress
- Work Log
- Work Log
- Work Log
- Work Log
- Skeleton deformation fitting
- Work Log
- Work Log
- GP Graph-matching for 3d reconstruction
- Work Log
- Work Log
- Diadem dataset investigation
- Work Log
- Brainstorm: Neuron-tracing
- Work Log
- Work Log
- New dynamic models
- Reconstruction problems
- Work Log
- Work Log
- Camera caliberation, revisited
- Thoughts on a New approach
- FIRE - Sampling strategy
- FIRE - improving fitting
- FIRE - Refactoring, k-folds cross validation, baseline evaluation
- FIRE - data prep, analyzing clustering
- Work Log
- FIRE - cleanup
- FIRE - first clustering test
- FIRE - cluster w/ missing data
- FIRE - streamlining; missing data
- FIRE Debugging cluster model
- FIRE = Cluster model
- FIRE - continuous model, clustering
- Fire: 10 inference experiments
- FIRE - piecewise linear inference
- Work Log
- Work Log
- Work Log
- Ellipse and Line Segment Detector ; Experiment Log
- Work Log
- Work Log
- TULIPS testing framework
- Test suite
- Work Log
- FIRE meeting notes
- FIRE date analysis
- Work Log
- Work Log
- Work Log
- Crash recovery; Index estimation - the saga continues
- TULIPS: index issue
- FIRE: Verifying date bug, export-to-csv
- Fire: mergin immunity data
- Debugging Index optimization
- Debugging reconstruction anomalies
- TULIPS - Debugging training
- Debugging training
- Experiment - Full-camera linearization
- Debugging WACV errors
- Work Log
- Debugging log
- FIRE - notes: joint meeting
- FIRE discussion
- FIRE - initial analysis
- FIRE - background reading, thinking and planning
- Proposal Practice Post-mortem
- Planning - Spring Interdisciplanry Computational Intelligence Seminar
- HPC meeting
- Dis Prop (ctd); new HPC
- Talk w/ Kobus Re: Dissertation Proposal
- Work Log
- Roughing out Proposal
- Dissertation Proposal - Preparation, Organization
- Reading: Semantic SLAM w/ GPLVM shape priors; FIRE reading
- Work Log
- Testing CL energy
- Work Log
- Work Log
- Optimizing indices: Length constraints
- Hyperprior
- Struggling with index offset and shrinkage
- Troubleshooting excessive index drift in endpoints; fixing Hessian under variable transformation.
- Work Log
- Work Log
- Index optimization, end-to-end
- Work Log
- Implementing, testing Hessian
- Covering Vision Course
- Work Log
- Chicken and Egg
- 2 Days: debugging WACV reconstruction
- Friday - iPlant Reading Group
- Debugging ML Gradient (part 2)
- Debugging ML gradient
- Work Log
- WACV Reconstruction issues; iPlant Reading; Index optimization
- Work Log
- WACV reconstruction (revisited)
- CVPR cleanup, documentation
- iPlant
- Post-CVPR-deadline; 2-part likelihood efficiency, 2-pass sampling
- (Two day) Markov Sampling (ctd). Implementing, testing, optimizing
- Optimizing posterior-sampling for pixel likelihood
- Ancestral Sampling from poserior (Markov-aware sampling)
- Implementing Nystrom method; bugs in posterior sampling code
- Testing likelihood #2 (2-day)
- Refactor: one-model-per-view likelihood
- Implementing Two-term likelihood
- GPU debugging
- Linux NVidia/Cuda/X11 erorrs; Cuda server; matlab integration
- KJB EM GMM
- Likelihood Training
- Work Log - Uninformative likelihood?
- Work Log - silhouettes, training likelihood, evaluating likelihood
- Work Log - Finishing Likelihood Server, integrating to sampling engine
- Work Log - Troubleshooting silhouette rendering
- Work Log - Four days of OpenGL Debugging
- Implementing Likelihood server
- New undertaking: C++ render server, pixel-based likelihood
- Split/merge toy (ctd)
- Problems with WACV Ground-truth reconstruction; Disasterous results with Split/merge toy problem
- Ground truth finished.
- Sampling - implemeting offline pair candidates
- Sampling - birth/death
- Discussion: maintaining attachment during sampling
- Misc maintenance & "Thinking" on sampling, priors
- refactoring; dependencies
- Testing full-tree covariance matrix
- Building reference implementation of full-tree covariance
- Full-tree covariance; Run on WACV dataset
- Branching prior covariance; implementing
- Branching prior covariance
- WACV results
- WACV Deadline
- Branching ML, debugging, training
- Misc.
- Attachment ML Math (ctd); Implementing Attach()/Detach()
- Clique-tree math (ctd)
- cleanup
- Branching curve clique-tree
- Saturday Thoughts - Enabling Non-gaussian models by using Gaussian models as proposal distributions
- Improved indexing; Retraining; Distinguishing between camera and plant motion
- Index Refinement; Mean-curve Reconstruction
- Background ML bugs; Why is Foreground Noise Variance so large?
- Theoretical Rate variance bug; Training background curve model
- Re-run training, Re-reconstruction, Curve-Flipping
- Pre-tails fix
- Refactoring, cleanup, bug fixes
- Visualizing Results; New training method
- Training, Reversed Curves, and Theoretical Rate Variance
- Singular Regions Issue; Training
- Training Bugs
- (No Title)
- (No Title)
- (No Title)
- , Week summary
- (No Title)
- (No Title)
- Maximum posterior with singular prior covariance
- (No Title)
- 2
- (No Title)
- (No Title)
- Direct Evaluation of the Marginal Likelihood
- (No Title)
- (No Title)
- (No Title)
- (No Title)
- (No Title)
- (No Title)
- Rethinking covariance functions
- (No Title)
- Foreground Curve Models as Gaussian Process Covariance Function
- Work summary
- (No Title)
- (No Title)
- (No Title)
- Rethinking Likelihood
- (No Title)
- (No Title)
- (No Title)
- Visualizing/Debugging BG ML

- Deriving likelihood of camera parameters
- Camera refinement
- Projecting Prior and Backprojecting Posterior
- Tree fitting - refactor
- Fitting the Deformation Model
- Articulated gaussian processes (part 2)
- Articulated Gaussian Processes
- Prior, visualized
- Trained prior parameters
- Projection of point onto line using distance from two reference points
- Geodesic distance kernel and BGP kernel -- simplified representation
- Graph matching with epipolar constraints
- GP with two constraints
- Generalizing the brownian bridge
- Neuron work planning
- FIRE - immunity data transformations
- FIRE - Self-report clustering results.
- FIRE immunity plots
- GPLVM & GPDM notes
- Constant-length energy function - Hessian
- Constant-length energy function -- revisited
- Constant-length energy function
- Marginal likelihood gradient (part 2)
- Hessian of Marginal Likelihood
- Gradient w.r.t. Indices
- Mixing Noisy and Noise-free values in GP Posterior
- params, CVPR 2014
- Track stages
- Split/Merge moves and Association Priors
- Summary of Dependency relationships