Searcharxiv⌕ Search

arXiv subjects

Lento Manickathan

Publications and source records attributed to Lento Manickathan.

3 recordsLinked to original sources

Pose-based Neural Clothing for Animated Characters

Traditional garment simulation for animated characters is a time-consuming bottleneck for artists due to slow simulations and tedious parameter tunings.Recent data-driven methods offer a powerful alternative, capable of learning from a dataset of simulated data to generate high-quality deformations. A lightweight network design allows data-driven systems to achieve a real-time inference speed. This significantly improves the artist workflow, as they can interactively pose a character and see high-quality folds and wrinkles at animation time, thereby allowing to make creative decisions ahead of time. However, existing neural clothing methods have presented appealing results on datasets based on realistic human models (\eg SMPL), but when trained on animated character datasets, they struggle to generalize to unseen poses. In this paper, we aim at designing neural networks and training strategies to generalize better to unseen poses in real-time to enable the interactive posing applications. We focus on a production-oriented setting where the model is trained for a single garment in a quasi-static setup. Technically, we leverage a smooth, expressive latent space by first training a Variational Auto-Encoder (VAE). We then progressively train the VAE encoder to transition from reconstruction to pose-driven prediction. To guide this prediction, we condition the network on both low-dimensional pose parameters and character body surface displacements. Finally, we introduce a novel local gradient histogram loss that mitigates the over-smoothing effect of reconstruction losses. Extensive evaluations on production-level assets and public datasets demonstrate that our method produces sharp, realistic wrinkles, smooth large-scale drape, and temporal coherence in real-time, offering a robust and easily integrable solution for interactive posing workflows.

cs.GR↗

Unveiling dynamic changes in the diurnal microclimate of a Buxus sempervirens with non-intrusive imaging of flow field, leaf temperature, and plant microstructure

Plant response is not only dependent on the atmospheric evaporative demand due to the combined effects of wind speed, air temperature, humidity, and solar radiation, but is also dependent on the water transport within the leaf-xylem-root system. Therefore, a detailed understanding of such dynamics is key to the development of appropriate mitigation strategies and numerical models. In this study, we unveil the diurnal dynamics of the microclimate of a Buxus sempervirens plant using multiple high-resolution non-intrusive imaging techniques. The wake flow field is measured using stereoscopic particle image velocimetry, the spatiotemporal leaf temperature history is obtained using infrared thermography, and additionally, the plant porosity is obtained using X-ray tomography. We find that the wake velocity statistics is not directly linked with the distribution of the porosity but depends mainly on the geometry of the plant foliage which generates the shear flow. The interaction between the shear regions and the upstream boundary layer profile is seen to have a dominant effect on the wake turbulent kinetic energy distribution. Furthermore, the leaf area density distribution has a direct impact on the short-wave radiative heat flux absorption inside the foliage where 50% of the radiation is absorbed in the top 20% of the foliage. This localized radiation absorption results in a high local leaf and air temperature. Furthermore, a comparison of the diurnal variation of leaf temperature and the net plant transpiration rate enabled us to quantify the diurnal hysteresis resulting from the stomatal response lag. The day of this plant is seen to comprise of four stages of climatic conditions: no-cooling, high-cooling, equilibrium, and decaying-cooling stages.

physics.flu-dyn↗

A hybrid Eulerian-Lagrangian flow solver

Currently, Eulerian flow solvers are very efficient in accurately resolving flow structures near solid boundaries. On the other hand, they tend to be diffusive and to dampen high-intensity vortical structures after a short distance away from solid boundaries. The use of high order methods and fine grids, although alleviating this problem, gives rise to large systems of equations that are expensive to solve. Lagrangian solvers, as the regularized vortex particle method, have shown to eliminate (in practice) the diffusion in the wake. As a drawback, the modelling of solid boundaries is less accurate, more complex and costly than with Eulerian solvers (due to the isotropy of its computational elements). Given the drawbacks and advantages of both Eulerian and Lagrangian solvers the combination of both methods, giving rise to a hybrid solver, is advantageous. The main idea behind the hybrid solver presented is the following. In a region close to solid boundaries the flow is solved with an Eulerian solver, where the full Navier-Stokes equations are solved (possibly with an arbitrary turbulence model or DNS, the limitations being the computational power and the physical properties of the flow), outside of that region the flow is solved with a vortex particle method. In this work we present this hybrid scheme and verify it numerically on known 2D benchmark cases: dipole flow, flow around a cylinder and flow around a stalled airfoil. The success in modelling these flow conditions presents this hybrid approach as a promising alternative, bridging the gap between highly resolved and computationally intensive Eulerian CFD simulations and fast but less resolved Lagrangian simulations.

math.NA↗