This webinar explores Rhino 8’s scripting environment, particularly the new Python 3 component, and how it enables direct integration of ML (Marchine Learning) models, such as regression or classification, for design optimization. While generative AI is often in the spotlight, many practical challenges in the AEC industry are better addressed with these models, which can significantly improve workflow efficiency.
One revisited project, LearnCarbon, showcases the potential of this integration. Initially, the project involved extensive data cleaning and model training using NumPy, Pandas, TensorFlow, and Keras, with predictions linked to Grasshopper via Hops. This workflow was dependent on external plugins and background scripts, leading to stability issues.
With Rhino 8’s new Python 3 integration, these models can now run directly within Rhino, eliminating the need for external plugins and improving workflow stability. This integration enables new possibilities:
- Accelerating slow processes by generating synthetic datasets in Grasshopper, or even utilizing existing ones, using ML to predict outcomes and bypassing traditional computationally-heavy simulations
- It allows for reversing model relationships—such as in LearnCarbon, where a model trained to predict embodied carbon based on materiality can now recommend optimal materials based on a target emission factor.
About the speaker:
Iliana Papadopoulou is an architect and software developer who has led initiatives focusing on spatial computing, machine learning, and interfaces. Her work spans all scales, from creating immersive virtual spaces to optimizing the performance of large urban layouts worldwide with neural networks. Iliana has contributed to both professional practice and academia, applying both research and technical skills to projects.
No comments:
Post a Comment