Datasets:
Update README.md
Browse files
README.md
CHANGED
|
@@ -40,6 +40,17 @@ For each surface, we included the following four files:
|
|
| 40 |
|
| 41 |
- __Blow-up Profile__ (`*_bup.txt`): A plain-text file containing the radius at each point along the discretized curve. These radii define the thickness of the tubular neighborhood used to generate the scalar field.
|
| 42 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 43 |
## Filename Encoding Convention
|
| 44 |
|
| 45 |
Filenames follow the pattern:
|
|
|
|
| 40 |
|
| 41 |
- __Blow-up Profile__ (`*_bup.txt`): A plain-text file containing the radius at each point along the discretized curve. These radii define the thickness of the tubular neighborhood used to generate the scalar field.
|
| 42 |
|
| 43 |
+
Each surface results from taking a tubular neighborhood and sinusoidal variations around Lissajous and Fibonacci parametrized curves.
|
| 44 |
+
|
| 45 |
+
The genus of each resulting surface is prescribed by the number of self-intersections of the generating curve. However, all elements in this dataset have been double-checked using the Euler characteristic function in Python’s trimesh module.
|
| 46 |
+
|
| 47 |
+
The watertightness of the mesh surfaces is ensured by the construction method, which uses the Marching Cubes algorithm; see details in our [paper](https://arxiv.org/abs/2505.13539) (Section 5).
|
| 48 |
+
|
| 49 |
+
## Dataset split
|
| 50 |
+
|
| 51 |
+
For training, we recommend preserving balanced genus distributions and splitting the dataset into 80% training and 20% testing. Each dataset sample consists of a group of files with a common filename encoding, as explained in the following section.
|
| 52 |
+
|
| 53 |
+
|
| 54 |
## Filename Encoding Convention
|
| 55 |
|
| 56 |
Filenames follow the pattern:
|