# CAD-Editor **Repository Path**: mirrors_microsoft/CAD-Editor ## Basic Information - **Project Name**: CAD-Editor - **Description**: No description available - **Primary Language**: Unknown - **License**: MIT - **Default Branch**: main - **Homepage**: None - **GVP Project**: No ## Statistics - **Stars**: 1 - **Forks**: 0 - **Created**: 2025-07-17 - **Last Updated**: 2026-02-21 ## Categories & Tags **Categories**: Uncategorized **Tags**: None ## README # CAD-Editor Official implementation of **[ICML 2025] CAD-Editor: A Locate-then-Infill Framework with Automated Training Data Synthesis for Text-Based CAD Editing** by *Yu Yuan, Shizhao Sun, Qi Liu, Jiang Bian*. 📄 [Paper](https://arxiv.org/abs/2502.03997) | 🤗 [Model](https://huggingface.co/microsoft/CAD-Editor) ｜ 🏠 [Project Page](https://cad-editor.github.io) ## Installation ```bash conda env create -f environment.yaml conda activate cad-editor ``` ## Data preparation We provide the complete data generation pipeline below for those who wish to generate their own dataset. We also share the data processed by us under `data/processed.zip`. ### 1. Paired CAD Generation **Step 1**: Generate design variations using hnc-cad. - Clone the [hnc-cad](https://github.com/samxuxiang/hnc-cad) repo. - Replace `gen/ac_gen.py` in the cloned hnc-cad repo with `hnc-cad/ac_gen.py` from this repo. Our updated version includes CAD model IDs (i.e., picture names) for pairing. - Follow the steps in the [hnc-cad](https://github.com/samxuxiang/hnc-cad) repo (especially `scripts/sample_cond.sh`) to generate design variations of a CAD model. **Step 2**: Convert generated `.obj` files to CAD sequences: ```python # Under utils folder: # Parse obj to primitive sequence python parse_obj2seq.py --input data \ --output data/dataset/train.pkl \ --bit 6 # Convert to our sequence format python convert.py --in_path data/dataset/train.pkl \ --out_path data/dataset/train_converted.json ``` **Step 3**: Pair CAD sequences: ```python python data/pair.py --in_path data/dataset/train_converted.json \ --out_path data/dataset/train_converted_pair.json ``` ### 2. Editing Instruction Generation **Visual Level** (1) Render CAD objects to images. ```python timeout 180 python utils/visual_obj.py --data_folder python utils/cad_img.py --input_dir \ --output_dir ``` (2) Generate captions. Please update the OpenAI endpoint information in `data/caption_image.py` before running. ```python python data/caption_image.py --sequence_dir data/dataset/train_converted_pair.json \ --image_dir data/dataset/train_img \ --caption_path data/dataset/train_caption_image.json ``` **Sequence Level** (1) Generate captions. ```python python data/caption_sequence.py --in_path data/dataset/train_converted_pair_2.json \ --out_path data/dataset/train_caption_sequence.json ``` ### 3. Merge and filter long-tailed sequences. ```python python data/merge.py --file1 data/dataset/train_caption_image.json \ --file2 data/dataset/train_caption_sequence.json \ --output data/dataset/train_all.json python data/filter_sequence.py --in_path data/dataset/train_all.json \ --out_path data/dataset/train.json ``` ## Training ### 1. Locating Stage **Step 1**: Create ground-truth masked CAD sequences: ```python # All training and inference are performed under the finetune folder: python create_mask.py --input_path \ --output_path ``` **Step 2**: Run locate training with multiple GPUs. Change `num_processes` in `ds_config.yaml` to specify how many GPUs will be used. ```python CUDA_VISIBLE_DEVICES= accelerate launch --config_file ds_config.yaml llama_finetune.py --task_type mask \ --run_name \ --data_folder \ --eval_freq 1000000 \ --save_freq 10000 ``` ### 2. Infilling Stage **Step 1**: Train infilling model: ```python CUDA_VISIBLE_DEVICES= accelerate launch --config_file ds_config.yaml finetune/llama_finetune.py --task_type infill \ --run_name \ --data_folder \ --eval_freq 1000000 \ --save_freq 10000 ``` **Step 2.** Enhanced training with selective data. Set `model_path` to the pretrained model from Step 1. Change `data_folder` to the folder of your selective data. ```python CUDA_VISIBLE_DEVICES= accelerate launch --config_file ds_config.yaml finetune/llama_finetune.py --task_type infill_selective \ --run_name \ --pretrained_model_path \ --data_folder \ --eval_freq 1000000 \ --save_freq 10000 ``` ## Inference Download our trained model checkpoints from [HuggingFace](https://huggingface.co/microsoft/CAD-Editor) to your ``````. ### 1. Locating Stage Generate masked sequences. Set the `` as ``. Set the `` as the path of `test.json` after unzip `data/processed.zip`. ```python CUDA_VISIBLE_DEVICES= python llama_sample.py \ --task_type mask \ --model_path \ --data_path \ --out_path \ --num_samples ``` ### 2. Infilling Stage Generate final edited CAD sequences. Set the `` as ``. Set the `` the same as the `out_path` of the locating stage. ```python CUDA_VISIBLE_DEVICES= python llama_sample.py \ --task_type infill \ --model_path \ --data_path \ --out_path \ --num_samples ``` ## Evaluation - Validity. ```python # Step 1: Parse the generated string to CAD obj. The in_path should be set the same as the out_path in the inference. python utils/parse_seq2obj.py --in_path \ --out_path \ --type infill # Step 2: Convert generated CAD obj to stl format. Use timeout command to prevent occ hanging. The data_folder should be set the same as the out_path in Step 1. timeout 180 python utils/visual_obj.py --data_folder # Step 3: Render and visualize to images. The input_dir should be set the same as the data_folder in Step 2. Use the number of successful generated images here to calculate the validity. python utils/cad_img.py --input_dir \ --output_dir ``` - 3D metrics (after running `visual_obj.py`). ```python # Under utils folder: # Uniformly sample points. Note that the generated CAD models and the ground truth test CAD models should be sampled respectively. python sample_points.py --in_dir \ --out_dir pcd # Evaluate performance. python eval_cad.py --fake \ --real ``` - Directional Clip Score ( Ensure you have run `cad_img.py` to render both the original and edited CAD sequences). ```python python eval_dclip.py --source_dir \ --edit_dir \ --instruction_path \ --out_path ``` ## Prompting-based Baselines We provide implementations of prompting-based baselines (including zero-shot and fewshot GPT-4o) under the `prompt/` folder. ## Citation If you find our work useful, please cite the following paper: ``` @article{yuan2025cad, title={CAD-Editor: A Locate-then-Infill Framework with Automated Training Data Synthesis for Text-Based CAD Editing}, author={Yuan, Yu and Sun, Shizhao and Liu, Qi and Bian, Jiang}, journal={Forty-Second International Conference on Machine Learning}, year={2025} } ``` ## Acknowledgements We would like to thank and acknowledge referenced codes from [hnc-cad](https://github.com/samxuxiang/hnc-cad), [SkexGen](https://github.com/samxuxiang/SkexGen) and [StyleGAN-nada](https://github.com/rinongal/StyleGAN-nada). ## Contributing This project welcomes contributions and suggestions. Most contributions require you to agree to a Contributor License Agreement (CLA) declaring that you have the right to, and actually do, grant us the rights to use your contribution. For details, visit https://cla.opensource.microsoft.com. When you submit a pull request, a CLA bot will automatically determine whether you need to provide a CLA and decorate the PR appropriately (e.g., status check, comment). Simply follow the instructions provided by the bot. You will only need to do this once across all repos using our CLA. This project has adopted the [Microsoft Open Source Code of Conduct](https://opensource.microsoft.com/codeofconduct/). For more information see the [Code of Conduct FAQ](https://opensource.microsoft.com/codeofconduct/faq/) or contact [opencode@microsoft.com](mailto:opencode@microsoft.com) with any additional questions or comments. ## Trademarks This project may contain trademarks or logos for projects, products, or services. Authorized use of Microsoft trademarks or logos is subject to and must follow [Microsoft's Trademark & Brand Guidelines](https://www.microsoft.com/en-us/legal/intellectualproperty/trademarks/usage/general). Use of Microsoft trademarks or logos in modified versions of this project must not cause confusion or imply Microsoft sponsorship. Any use of third-party trademarks or logos are subject to those third-party's policies.