|
|
--- |
|
|
license: cc-by-4.0 |
|
|
task_categories: |
|
|
- text-to-3d |
|
|
- other |
|
|
language: |
|
|
- en |
|
|
size_categories: |
|
|
- 100K<n<1M |
|
|
tags: |
|
|
- cad-function |
|
|
- text-to-cad |
|
|
- dataset |
|
|
- annotation |
|
|
--- |
|
|
|
|
|
# Req2CAD: Bridging Functional Requirements and Parametric CAD Models to Support Conceptual 3D Design |
|
|
|
|
|
## ๐ Dataset Summary |
|
|
|
|
|
The **Req2CAD** dataset is constructed on top of the **Text2CAD** dataset, which itself extends **DeepCAD** with textual annotations and rendered images. Building upon these resources, Req2CAD further enriches each CAD component with **functional annotations**, establishing a direct mapping between functional requirements and structural design representations. |
|
|
|
|
|
Each sample contains the following core fields: |
|
|
|
|
|
- **uid**: Unique identifier, consistent with DeepCAD and Text2CAD |
|
|
- **function_description**: Natural language descriptions of CAD component functions |
|
|
- **function_keywords**: Structured function labels in the form of *verb + object*, extracted from function descriptions |
|
|
|
|
|
The function descriptions are generated by a VLM, **Qwen2.5-VL-7B-Instruct**, based on the input multi-view rendered images of CAD components, and the function keywords are extracted from the function descriptions by an LLM, **Qwen3-14B**. |
|
|
|
|
|
The multi-view rendered images of CAD components can be downloaded from the [Text2CAD](https://huggingface.co/datasets/SadilKhan/Text2CAD) dataset, and the parametric CAD representations and construction sequences can be downloaded from the [DeepCAD](https://github.com/rundiwu/DeepCAD) dataset. |
|
|
|
|
|
## ๐ License |
|
|
|
|
|
This dataset is released under the **CC-BY-4.0** license. |
|
|
|
|
|
<!-- --- |
|
|
|
|
|
## ๐ Citation |
|
|
|
|
|
If you use this dataset, please cite: |
|
|
|
|
|
```bibtex |
|
|
@dataset{jing2026_fskb, |
|
|
title = {Function--Structure Knowledge Base for Conceptual CAD Design}, |
|
|
author = {Jing, Qianzhi and collaborators}, |
|
|
year = {2026}, |
|
|
publisher = {Hugging Face} |
|
|
} |
|
|
--> |