![[PROJECT TITLE]](../res/img/synthetic-720.webp)
Training computer vision models for defect detection in manufacturing requires extensive data, which is costly and time-consuming to obtain in the real world. This project presents a digital Manufacturing methodology to generate high-fidelity synthetic data simulating fiber optic faults using virtual reality environments and circular economy models.
Utilizes advanced rendering tools to create images that accurately replicate physical fiber optic defects.
Validates synthetic images using Mean Structural Roughness (Ms), Mean Color Variance (Mc), and Fréchet Inception Distance (FID), ensuring realism and variability.
Evaluates the impact of using virtual versus physical data through a multidimensional index encompassing economic, environmental, social, and technological factors aligned with sustainability.
Detailed capabilities and technical specifications developed for this mission.
Minimizes the need to collect physical defective samples, significantly reducing resource consumption, carbon footprint, and material waste.
Aligns the synthetic data lifecycle with product development phases (CADMID) and circular economy strategies (R10), promoting sustainable manufacturing practices.
Enables the controlled introduction of variations and the creation of rare defect scenarios that are difficult to capture under real conditions, improving the robustness of predictive models.
Facilitates synchronization between physical processes and their virtual representations, enabling dynamic adjustments and advanced diagnostics without interrupting production operations.