Live Demonstration: Defect Detection in Material Extrusion with Multimodal Large Language Model

Defect detection in material extrusion (MEX) is crucial because defects can significantly impact the functionality and mechanical properties of 3D-printed parts. Defect detection methods based on convolutional neural networks (CNNs) often involve extensive labelled datasets and substantial computational resources [1]. On the other hand, methods that focus on identifying discrepancies between the 3D-printed part and the intended design cannot classify specific types of defects [2]. This demonstration introduces Multimodal Large Language Models (MLLMs), which integrate textual and visual data, minimizing the need for large datasets and providing a more efficient approach to defect detection in MEX.