Official python implementation for the paper: Contrastive Learning-based Representation Learning for Computer-Aided Design Models
(Updated 04-30-2024) Now you can use Docker to train & test our model, without installing all the dependencies locally, by running: docker run --gpus all -it --rm fmsjung/contrastcad
You can start training ContrastCAD with desired number of epoch by running below command (-g 0 is for GPU id):
python train_cl.py --exp_name contrastcad -g 0
If you would like to use out RRE data augmentation, simply attach some arguments as below:
python train_cl.py --exp_name contrastcad -g 0 --augment --dataset_augment_type rre
Once your ContrastCAD is done training, you may train latent-GAN as latent generative model for your ContrastCAD:
# encode data
python test_cl.py --exp_name contrastcad --mode enc --ckpt {epoch-number-here} -g 0
# train lgan
python lgan.py --exp_name contrastcad --ae_ckpt {epoch-number-here} -g 0
To evaluate the reconstruction performance:
# reconstruct data
python test_cl.py --exp_name contrastcad --mode rec --ckpt {epoch-number-here} -g 0
# evaluate
cd evaluation
# for accuracy
python evaluate_ae_acc.py --src ../proj_log/contrastcad/results/test_{epoch_number-here}
# for CD and invalid rate
python evaluate_ae_cd.py --src ../proj_log/contrastcad/results/test_{epoch_number-here} --parallel
If you would like to use our pretrained checkpoints, simply replace exp_name arguemtn with pretrained such as :
python test_cl.py --exp_name pretrained --mode rec --ckpt {epoch-number-here} -g 0
For dataset, please refer to DeepCAD repository. Download dataset and unzip them into data directory (or whatever directory name you specify).
@ARTICLE{10559801,
author={Jung, Minseop and Kim, Minseong and Kim, Jibum},
journal={IEEE Access},
title={ContrastCAD: Contrastive Learning-Based Representation Learning for Computer-Aided Design Models},
year={2024},
volume={12},
number={},
pages={84830-84842},
keywords={Solid modeling;Shape measurement;Computational modeling;Data models;Training;Three-dimensional displays;Transformers;Design automation;Contrastive learning;CAD model;transformer autoencoder;CAD generation},
doi={10.1109/ACCESS.2024.3415816}}
React
Typescript
Django
Laravel
Facebook
Microsoft
Google
Alibaba
D3
Tencent