Konversatorium on Friday, October 5, 2018 - 10:30
Toward Whole Cell Mesoscale Models
The ability to create structural models of cells and cellular components at the molecular level is being driven by advances ranging from proteomics and expression profiling to ever more powerful imaging approaches. It is being enabled by technology leaps in computation, informatics, and visualization.
Our CellPACK program is a tool for generating structural models of cellular environments at molecular and atomic level. Recently we have implemented a GPU-based implementation of CellPACK that speeds up the process by orders of magnitude, in what we have termed “instant packing.” This enables interactive exploration and manipulation of components of the packings. Visualization of these models is also enabled by GPU-based efficient representations, renderings and levels of detail. The ability to model complex cellular components such as a bacterial nucleoid and distinct phases is a significant challenge that we continue to work on. Our Lab’s recent lattice-based method for rapidly producing bacterial nucleoids is a prototype for other rule-based generative structure builders. Use of GPU-based physics engines enables real time interaction with dynamic models. Flexx is a real-time constraint solver from NVidia. It is capable of interactive constraint minimization of up to 1 million particles in real time. Such systems can quickly resolve clashes and identify interactions in the crowded cellular environment.
In a parallel effort to CellPACK, we have developed CellPAINT, which has its origins in David Goodsell’s watercolor paintings of cellular environments and processes. Cellpaint uses a painting metaphore to enable creation of Goodsell-like images interactively within a Unity game-engine. These images can be animated using a simple Brownian motion-based diffusion model. Recently we have expanded this interactive interface into 3D, and have also implemented it in a Virtual Reality environment.
The talk will include live interactive demonstrations of the current state of our software.
StretchDenoise: Parametric Curve Reconstruction with Guarantees by Separating Connectivity from Residual Uncertainty of Samples (Pacific Graphics test talk)
We reconstruct a closed denoised curve from an unstructured and highly noisy 2D point cloud. Our proposed method uses a two-pass approach: Previously recovered manifold connectivity is used for ordering noisy samples along this manifold and express these as residuals in order to enable parametric denoising. This separates recovering low-frequency features from denoising high frequencies, which avoids over-smoothing. The noise probability density functions (PDFs) at samples are either taken from sensor noise models or from estimates of the connectivity recovered in the first pass. The output curve balances the signed distances (inside/outside) to the samples. Additionally, the angles between edges of the polygon representing the connectivity become minimized in the least-square sense. The movement of the polygon’s vertices is restricted to their noise extent, i.e., a cut-off distance corresponding to a maximum variance of the PDFs. We approximate the resulting optimization model, which consists of higher-order functions, by a linear model with good correspondence. Our algorithm is parameter-free and operates fast on the local neighborhoods determined by the connectivity. This enables us to guarantee stochastic error bounds for sampled curves corrupted by noise, e.g., silhouettes from sensed data, and we improve on the reconstruction errorfrom ground truth. Source code is available online. An extended version is available at: https://arxiv.org/abs/1808.07778
Sit & Relax: Interactive Design of Body-Supporting Surfaces (Pacific Graphics test talk)
We propose a novel method for interactive design of well-fitting body-supporting surfaces that is driven by the pressure distribution on the body's surface.
Our main contribution is an interactive modeling system that utilizes captured body poses and computes an importance field that is proportional to the pressure distribution on the body for a given pose. This distribution indicates where the body should be supported in order to easily hold a particular pose, which is one of the measures of comfortable sitting.
Using our approximation, we propose the entire workflow for interactive design of $C^2$ smooth surfaces which serve as seats, or generally, as body supporting furniture for comfortable sitting. Finally, we also provide a design tool for Rhino/Grasshopper that allows for interactive creation of single designs or entire multi-person sitting scenarios. We also test the tool with design students and present several results.
Our method aims at interactive design in order to help designers to create appropriate surfaces digitally without additional empirical design passes.