Mesh2Sequences

Automated Conversion of Unstructured Mesh Models to Structured Parametric CAD Models

Flow Chart

Environment Setup

Launch Fusion 360 Environment

1. Install Fusion 360 (requires educational or commercial account)
2. Create a conda virtual environment:

   conda create -n mesh2sequences python=3.9
   conda activate mesh2sequences

3. Install Python dependencies:

   pip install -r requirements.txt

4. Launch Gym server in Fusion 360:
   • Open Fusion 360
   • Go to Tools > Add-ins > Scripts and Add-ins
   • Add and run the Fusion 360 Gym server

Training

Navigate to train/src directory and run the training script:

cd train/src
python train.py --dataset /path/to/data/ --split /path/to/train_test.json

Main arguments:

• --dataset: Path to dataset
• --split: Train/test split file
• --mpn: Network type, options: gcn, mlp [default: gcn]
• --epochs: Number of training epochs [default: 100]
• --lr: Learning rate [default: 0.0001]

Trained models are saved in train/ckpt/ directory.

Download Model Weights

Trained model weights are available on Google Drive. Download the model checkpoints and place them in the train/ckpt/ directory for inference.

Inference

Navigate to inference directory and run the inference script:

cd inference
python main.py --input /path/to/target.step --agent gcn --search best

Main arguments:

• --input: Target B-Rep file or folder
• --agent: Agent type, options: gcn, mlp [default: gcn]
• --search: Search strategy, options: beam, best [default: best]
• --budget: Number of search steps [default: 100]
• --launch_gym: Auto-launch Fusion 360 Gym [default: False]

Results are saved in inference/log/ directory.