adjust width
Browse files
README.md
CHANGED
|
@@ -11,7 +11,10 @@ This dataset contains 150,000 procedurally synthesized 3D shapes in order to hel
|
|
| 11 |
|
| 12 |
Our dataset is collected based on recent works [Xie et al. (2024)](https://desaixie.github.io/lrm-zero/), and we utilized procedure generated data in self-supervised setting. Each 3D shape is represented by uniformly sampled surface points, making it a versatile resource for pretraining models for tasks such as masked point cloud completion, shape classification, and more.
|
| 13 |
|
| 14 |
-
|
|
|
|
|
|
|
|
|
|
| 15 |
*Figure 1. Examples of procedurally generated 3D shapes showcasing varying geometric complexity. In this dataset, we only provide data in the category of (d). Please checkout github if you want to render data in different complexity level.*
|
| 16 |
|
| 17 |
## Key Features
|
|
@@ -28,7 +31,10 @@ Our dataset is collected based on recent works [Xie et al. (2024)](https://desai
|
|
| 28 |
## Dataset Size and Performance
|
| 29 |
We evaluated the impact of dataset size on the **PB-T50-RS benchmark** for shape classification using Point-MAE-Zero. Our findings show that performance improves with larger dataset sizes but exhibits diminishing returns beyond a certain threshold.
|
| 30 |
|
| 31 |
-
|
|
|
|
|
|
|
|
|
|
| 32 |
*Figure 2. The effect of dataset size on downstream shape classification performance. Note that our performance is on par with Point-MAE trained with ShapeNet at exactly the same scale.*
|
| 33 |
|
| 34 |
Additional experiments are available in [our paper](https://arxiv.org/abs/xxxx.xxxxx).
|
|
|
|
| 11 |
|
| 12 |
Our dataset is collected based on recent works [Xie et al. (2024)](https://desaixie.github.io/lrm-zero/), and we utilized procedure generated data in self-supervised setting. Each 3D shape is represented by uniformly sampled surface points, making it a versatile resource for pretraining models for tasks such as masked point cloud completion, shape classification, and more.
|
| 13 |
|
| 14 |
+
|
| 15 |
+
<p align="center">
|
| 16 |
+
<img src="./figs/dataset-complexity.jpg" alt="Shape Complexity" style="width: 80%;">
|
| 17 |
+
</p>
|
| 18 |
*Figure 1. Examples of procedurally generated 3D shapes showcasing varying geometric complexity. In this dataset, we only provide data in the category of (d). Please checkout github if you want to render data in different complexity level.*
|
| 19 |
|
| 20 |
## Key Features
|
|
|
|
| 31 |
## Dataset Size and Performance
|
| 32 |
We evaluated the impact of dataset size on the **PB-T50-RS benchmark** for shape classification using Point-MAE-Zero. Our findings show that performance improves with larger dataset sizes but exhibits diminishing returns beyond a certain threshold.
|
| 33 |
|
| 34 |
+
|
| 35 |
+
<p align="center">
|
| 36 |
+
<img src="./figs/scaling_law.jpg" alt="Impact of Dataset Size" style="width: 80%;">
|
| 37 |
+
</p>
|
| 38 |
*Figure 2. The effect of dataset size on downstream shape classification performance. Note that our performance is on par with Point-MAE trained with ShapeNet at exactly the same scale.*
|
| 39 |
|
| 40 |
Additional experiments are available in [our paper](https://arxiv.org/abs/xxxx.xxxxx).
|