Related Experiment Video
Updated: Jun 26, 2026
06:53
Additive Manufacturing of Functionally Graded Ceramic Materials by Stereolithography
Published on: January 25, 2019
CeraMIRScan: Mid-infrared OCT Scan Dataset for Ceramic Quality Assessment
Natalia P García-de-la-Puente1, Fernando García-Torres2, Andrés Laveda-Martínez3
1Instituto Universitario de Investigación en Tecnología Centrada en el Ser Humano, HUMAN-tech, Universitat Politècnica de València, Valencia, Spain. napegar@upv.es.
Scientific Data
|June 24, 2026
Summary
A new dataset, CeraMIRScan, aids ceramic quality assessment using Mid-infrared Optical Coherence Tomography (MIR-OCT) and deep learning. It enables automated defect detection and segmentation in industrial applications.
Area of Science:
- Materials Science
- Non-Destructive Testing (NDT)
- Artificial Intelligence
Background:
- Mid-infrared Optical Coherence Tomography (MIR-OCT) offers high-resolution imaging for industrial NDT.
- Deep Learning (DL) models for defect detection in MIR-OCT scans, especially for ceramics, are underdeveloped.
- Automated quality assessment of 3D printed ceramics using MIR-OCT is an emerging area.
Purpose of the Study:
- Introduce the CeraMIRScan dataset for ceramic quality assessment.
- Provide a benchmark for developing DL models for defect detection and segmentation in MIR-OCT scans.
- Facilitate advancements in automated quality control for ceramic manufacturing.
Main Methods:
- Developed the CeraMIRScan dataset comprising 29 MIR-OCT volumes (21,882 B-scans) of 3D printed ceramics.
- Expert-annotated binary masks identify defects like pores, delaminations, and inclusions.
- Implemented a U-Net architecture for baseline DL-based defect segmentation.
Main Results:
- The CeraMIRScan dataset contains 41.38% of images with visible anomalies.
- Baseline U-Net model achieved 80.55% precision, 80.00% recall, and 80.27% Dice score for defect segmentation.
- Demonstrated the dataset's utility for training DL models for ceramic defect analysis.
Conclusions:
- CeraMIRScan serves as a valuable resource for advancing automated quality assessment in ceramics.
- The dataset supports the development of MIR-OCT-based defect characterization techniques.
- Highlights the potential of DL for enhancing NDT in industrial ceramic applications.
