nugets
NUGETS — NeUral GEomeTry benchmarkS
NUGETS is a benchmark suite for neural geometry processing tasks.
It provides a comprehensive set of datasets and tasks for evaluating the performance of neural networks on geometry tasks.
Methodology:
- Datasets:
We provide a set of synthetic and real-world datasets for evaluating neural networks on geometry tasks.
- Tasks:
For each dataset, we provide a set of tasks that can be used to evaluate the performance of neural networks on geometry tasks.
- Metrics:
We also provide various choices for metrics one can track during training e.g. mean squared error, relative error, more specialised error metrics that are task specific. These can be found in the losses.
Modules
Geometric datasets |
|
Loss functions |
|
Miscellaneous utilities |
|
Training and evaluation pipeline on geometric data. |