LZ
L.K. Zimmerhackl
info
Please Note
<p>This page displays the records of the person named above and is not linked to a unique person identifier. This record may need to be merged to a profile.</p>
2 records found
1
In this thesis, we address the problem of learning mesh-specific, impulse-dependent fracture patterns in real time. Our approach is based on regressing a distance field over the mesh surface, encoding the proximity of each vertex to fracture lines, which is subsequently segmented into distinct pieces using graph-based methods such as watershed segmentation. The goal is to achieve real-time performance, which is something the current approach does not achieve for large meshes.
We evaluate different neural architectures, comparing a multilayer perceptron to DeltaConv, a graph convolutional model, and find that the MLP provides superior performance. In addition, we assess multiple segmentation strategies and identify watershed as the most effective, followed by hierarchical segmentation. We also find that the segmentation algorithms do not achieve real-time performance for large meshes.
These results highlight the potential of machine learning-based fracture simulations, but also indicate that distance field segmentation is not capable of real-time performance using our tested algorithms. This suggests that future work should focus on directly learning the labels rather than relying on distance fields as an intermediary representation in real-time scenarios.
...
We evaluate different neural architectures, comparing a multilayer perceptron to DeltaConv, a graph convolutional model, and find that the MLP provides superior performance. In addition, we assess multiple segmentation strategies and identify watershed as the most effective, followed by hierarchical segmentation. We also find that the segmentation algorithms do not achieve real-time performance for large meshes.
These results highlight the potential of machine learning-based fracture simulations, but also indicate that distance field segmentation is not capable of real-time performance using our tested algorithms. This suggests that future work should focus on directly learning the labels rather than relying on distance fields as an intermediary representation in real-time scenarios.
...
In this thesis, we address the problem of learning mesh-specific, impulse-dependent fracture patterns in real time. Our approach is based on regressing a distance field over the mesh surface, encoding the proximity of each vertex to fracture lines, which is subsequently segmented into distinct pieces using graph-based methods such as watershed segmentation. The goal is to achieve real-time performance, which is something the current approach does not achieve for large meshes.
We evaluate different neural architectures, comparing a multilayer perceptron to DeltaConv, a graph convolutional model, and find that the MLP provides superior performance. In addition, we assess multiple segmentation strategies and identify watershed as the most effective, followed by hierarchical segmentation. We also find that the segmentation algorithms do not achieve real-time performance for large meshes.
These results highlight the potential of machine learning-based fracture simulations, but also indicate that distance field segmentation is not capable of real-time performance using our tested algorithms. This suggests that future work should focus on directly learning the labels rather than relying on distance fields as an intermediary representation in real-time scenarios.
We evaluate different neural architectures, comparing a multilayer perceptron to DeltaConv, a graph convolutional model, and find that the MLP provides superior performance. In addition, we assess multiple segmentation strategies and identify watershed as the most effective, followed by hierarchical segmentation. We also find that the segmentation algorithms do not achieve real-time performance for large meshes.
These results highlight the potential of machine learning-based fracture simulations, but also indicate that distance field segmentation is not capable of real-time performance using our tested algorithms. This suggests that future work should focus on directly learning the labels rather than relying on distance fields as an intermediary representation in real-time scenarios.
Dependent programming languages such as Agda show a lot of promise in creating new ways of writing code, but currently suffer from a lack of support and features. In this paper we attempt to create a new back-end for Agda targeting Java which has a huge and thriving ecosystem.
We implement the new back-end for Agda in Haskell and we describe the benefits and drawbacks of targeting Java. Firstly we go into the existing methods of compiler Agda, then we go into how to compile Agda to Java and what the main challenges where creating the compiler and what solutions were implemented to solve these. Afterwards Agda2Java is compared to the existing methods of compiling Agda code by means of benchmarks and analyzing the execution time. We show that at its current state, Java does not seem to be a promising back-end for Agda, but that there is work being done on Java that might change this perception. ...
We implement the new back-end for Agda in Haskell and we describe the benefits and drawbacks of targeting Java. Firstly we go into the existing methods of compiler Agda, then we go into how to compile Agda to Java and what the main challenges where creating the compiler and what solutions were implemented to solve these. Afterwards Agda2Java is compared to the existing methods of compiling Agda code by means of benchmarks and analyzing the execution time. We show that at its current state, Java does not seem to be a promising back-end for Agda, but that there is work being done on Java that might change this perception. ...
Dependent programming languages such as Agda show a lot of promise in creating new ways of writing code, but currently suffer from a lack of support and features. In this paper we attempt to create a new back-end for Agda targeting Java which has a huge and thriving ecosystem.
We implement the new back-end for Agda in Haskell and we describe the benefits and drawbacks of targeting Java. Firstly we go into the existing methods of compiler Agda, then we go into how to compile Agda to Java and what the main challenges where creating the compiler and what solutions were implemented to solve these. Afterwards Agda2Java is compared to the existing methods of compiling Agda code by means of benchmarks and analyzing the execution time. We show that at its current state, Java does not seem to be a promising back-end for Agda, but that there is work being done on Java that might change this perception.
We implement the new back-end for Agda in Haskell and we describe the benefits and drawbacks of targeting Java. Firstly we go into the existing methods of compiler Agda, then we go into how to compile Agda to Java and what the main challenges where creating the compiler and what solutions were implemented to solve these. Afterwards Agda2Java is compared to the existing methods of compiling Agda code by means of benchmarks and analyzing the execution time. We show that at its current state, Java does not seem to be a promising back-end for Agda, but that there is work being done on Java that might change this perception.