DEEP Inspection for Materials Science¶

Inspection and characterization of materials are fundamental for understanding material properties, ensuring quality control, and accelerating materials development and advanced manufacturing. However, conventional approaches often require extensive expert analysis of complex images, making them time-consuming and difficult to scale. With the rapid growth of high-resolution imaging techniques, deep learning-based methods provide powerful solutions for automated materials image analysis by learning meaningful visual features for tasks such as defect classification, microstructure recognition, process monitoring, and intelligent quality assessment.

This workshop offers a focused, hands-on introduction to deep learning for visual inspection in materials science. Using the Severstal Steel Defect Detection dataset as a concrete, running example, participants will follow a complete analysis pipeline, from raw image data through model training to interpretable results. The techniques covered are deliberately chosen for their breadth of applicability: while the examples are grounded in industrial quality control, the same approaches translate directly to defect detection in electron microscopy, anomaly identification in medical imaging, and structural characterization across a wide range of domains.

Prerequisites

Basic Python programming (loops, functions, libraries)
Familiarity with NumPy or similar data manipulation tools is helpful but not required
No prior deep learning experience necessary – core concepts will be introduced from scratch

Setup

Episodes

Learning outcomes¶

This material is for practitioners and researchers in materials science and industrial inspection who want to apply deep learning to surface defect analysis. No prior deep learning experience is required, but familiarity with Python is assumed. The techniques covered transfer directly to related domains such as electron microscopy, medical imaging, crystallography, and broader manufacturing quality control.

By the end of this workshop, learners should be able to:

Explain the two core assumptions behind CNNs (translation invariance and locality) and why they make convolution more parameter-efficient than fully connected layers
Describe and distinguish four computer vision paradigms: anomaly detection, image classification, object detection, and image segmentation, and identify which is appropriate for a given inspection task
Apply transfer learning (feature extraction and fine-tuning) with pretrained CNN and transformer architectures (VGG, ResNet, ViT) to a new defect classification dataset
Explain how PatchCore detects anomalies without labeled defect examples, and contrast this with supervised classifiers and their open-set failure modes
Understand the YOLO family’s progression from two-stage pipelines to single-pass anchor-free detection, and apply YOLO11 to both object detection and instance segmentation tasks
Distinguish semantic segmentation (U-Net, SegFormer) from instance segmentation (YOLO11-seg) and select the appropriate architecture for a given materials inspection scenario

Credit¶

This material was developed by the Sweden AI Factory at NAISS/NSC, Linköping University and RISE.

Parts of the curriculum are adapted from Dive into Deep Learning (CC BY-SA 4.0).

Use of AI assistants

Parts of this material — including draft text, code cells and example notebooks — were produced with the help of AI coding assistants (e.g. Anthropic Claude, GitHub Copilot). All content has been reviewed, tested and edited by the authors, who take full responsibility for its accuracy.

License¶

Licensing

Unless noted otherwise, this material is dual-licensed:

Teaching material, text, figures and other media — CC BY-SA 4.0
Source code, code cells and snippets — MIT

The full license texts are in LICENSE and LICENSE.code in the repository.

Suggested attribution

“DEEP Inspection for Materials Science” by Sweden AI Factory, used under CC BY-SA 4.0.

CC BY-SA 4.0 — what this means

You are free to

Share — copy and redistribute the material in any medium or format, for any purpose, even commercially.
Adapt — remix, transform and build upon the material, for any purpose, even commercially.

The licensor cannot revoke these freedoms as long as you follow the license terms.

Under the following terms

Attribution — you must give appropriate credit, provide a link to the license, and indicate if changes were made.
ShareAlike — if you remix or build upon the material, you must distribute your contributions under the same license.
No additional restrictions — you may not apply legal terms or technological measures that legally restrict others from doing anything the license permits.

Notices

You do not have to comply with the license for elements of the material in the public domain, or where your use is permitted by an applicable exception or limitation. No warranties are given; other rights such as publicity, privacy or moral rights may limit how you use the material.

This is a human-readable summary, not a substitute for the legal code.

MIT License — full text

Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation files (the “Software”), to deal in the Software without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, and to permit persons to whom the Software is furnished to do so, subject to the following conditions:

The above copyright notice and this permission notice shall be included in all copies or substantial portions of the Software.

THE SOFTWARE IS PROVIDED “AS IS”, WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.