Revolutionizing Architectural Annotations

ServCAD is an innovative software tool designed to enhance architectural modeling by incorporating advanced annotation and prediction capabilities. Utilizing machine learning, particularly Graph Neural Networks (GNNs), ServCAD provides a robust platform for both generating and processing architectural drawings. Below is an overview of how ServCAD operates in the training and testing phases.

Training Phase

1.Data Generation:

• Input: Architectural drawings with annotations.

• Process: The drawings are input into ServCAD, which processes these inputs to generate structured data files (such as element_data.csv and Elementinfo.txt).

2.Graph Generation:

• From the data generated, a graph is constructed to represent the relationships and dependencies among various architectural elements.

3.Label Prediction via GNN:

• Using the graph data, a Graph Neural Network is trained to predict labels for various elements, enhancing the model's ability to automatically annotate and recognize components in architectural designs.

Testing Phase

1.Model Application:

• Input: A raw model without annotations.

• Process: ServCAD applies the trained GNN model to the raw input, predicting annotations for the elements within the model.

2.Output Generation:

• The output is an annotated model, where each element has been labeled according to the predictions made by the GNN.

• This model is saved alongside the element_data.csv which now includes predicted label types.

ServCAD streamlines the process of architectural design by automating the annotation and component recognition tasks, reducing manual input and increasing accuracy and efficiency. The integration of GNN for predictions allows for continuously improving accuracy as more annotated data is fed into the system during the training phase.