Reproduce “Towards Smart Monitored AM: Open Source in-Situ Layer-wise 3D Printing Image Anomaly Detection Using Histograms of Oriented Gradients and a Physics-Based Rendering Engine” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Aliaksei Petsiuk, Joshua M. Pearce
Researchers developed an open-source, layer-wise anomaly detection system for 3D printing that uses a physics-based rendering engine to generate synthetic reference images. This method, combined with Histograms of Oriented Gradients (HOG) feature comparison, reliably segments diverse print defects with a minimum resolution of 5 mm without requiring training data.
Key insights
- The integration of a physics-based rendering engine significantly enhances the fidelity of reference images, allowing for sophisticated image processing and precise anomaly detection.
- HOG-based similarity comparison is efficient enough for in-situ layer-wise monitoring, with the primary time cost being the one-time preliminary rendering of synthetic images per part geometry.
- Specific similarity metrics, such as Kendall’s tau, Jaccard, and Sorensen, demonstrate high sensitivity to deviations, while Pearson's r, Spearman's rho, cosine, and Dice similarities offer reliable results for various defect types.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.