GTSAM documentation
Release overview figures
Release overview figures
The figures support the technical summaries on the landing page.
landing-page-snapshot.jpg is a 1280 × 720 browser screenshot of the landing
page, captured from the local website preview on September 23, 2026. It is used
in the 4.3 release announcement and downscaled for that post’s blog thumbnail.
It is captured separately, not generated by build_release_figures.py.
plaza2-notebook.png: static rendering of the final saved Plotly figure in RangeSLAMExample_plaza2.ipynb, comparing the initial odometry trajectory, optimized path, and landmarks. Data and styling are unchanged; exported at 900 × 650 CSS pixels with 2× resolution. The interactive notebook is linked from the landing-page figure.pose2-notebook.png: unchanged PNG output embedded in Pose2SLAMExample.ipynb, plotting optimized poses and marginal covariances.hybrid-bayes-tree.svg: unchanged data-association tutorial figure. The original embeds its fonts.cuda-speedups.svgandcuda-speedups-mobile.svg: desktop and phone layouts of the reported values in the CUDA article. General-path values are maxima across benchmarks; GPU-resident SfM values refer to three specific BAL problems. All timings are complete optimizer wall times on an NVIDIA A100. The reported ratios are used directly, not recomputed from rounded seconds.multifrontal-chain.svg: normal matrix and upper Cholesky factor computed from the unit-noise four-variable chain in MultifrontalSolver.ipynb. This is a numerical illustration, not a benchmark or output of the C++ solver.
Reproduce these files using the py312 environment (including Plotly and Kaleido):
python scripts/build_release_figures.py --gtsam-repo /path/to/gtsam
The script pins extraction to commit
97b7a5e3b8edce204320397f00b502f57952fb38 (GTSAM 4.3.0).
The copied GTSAM outputs retain the accompanying LICENSE.BSD notice.
The five pre-existing research figures and the legged-estimation animation are
reused in place and credited in the page captions.