SECAD-Net

This repository provides the official code of SECAD-Net.

📢: We will release the code for evaluation and the preprocessed Fusion360 dataset later. For detailed information or any inquiries, please contact pli142857@gmail.com.

📢: The current code can only use marching cubes to generate mesh, and the code to generate CAD format is currently a stub function. I have proposed an initial implementation in this issue discussion. After thorough testing and validation, I intend to merge this code into the main codebase of the repository.

Dependencies

Install python package dependencies:

$ pip install -r requirements.txt

Dataset

We used the ABC dataset processed by CAPRI-Net, please download it from the link abc_all.zip they provided.

Training & Fine-tuning & Testing

We provide you with basic experiment scripts:

$ sh scripts/train.sh
$ sh scripts/fine-tuning.sh
$ sh scripts/test.sh

Pre-trained model

Download the pretrained model from this link and put it under exp_log/ABC/ModelParameters/, then run the fine-tuning code.

Acknowledgement

We would like to thank and acknowledge referenced codes from CSGStumpNet, CAPRI-Net and DeepCAD.