Anomaly Detection Template Kit (DEMO)

This is a hands-on, production-oriented AI Template Kit for anomaly detection in repetitive visual structures such as fabrics, surfaces, and materials.
It provides fully working, modular implementations using VAE, PatchCore, and DRAEM — structured for direct use in real projects.

While each method is thoroughly documented in Jupyter Notebooks to support learning, the primary focus is on deployable, real-world anomaly detection pipelines — not academic exploration.

---

🎁 About This Demo Version

This is a limited demo of the full Anomaly Detection Template Kit.

It includes the complete VAE track and preview notebooks for PatchCore and DRAEM, giving you a clear impression of the structure, modularity, and quality of the full package.

---

🔓 What’s Included in the Demo

• ✅ Full source code and notebooks for VAE-based anomaly detection
• ✅ Two preview notebooks introducing PatchCore and DRAEM
• ✅ Dataset preparation utilities and basic evaluation workflows

---

🔐 What’s in the Full Version

The full version is a professional-grade template kit that includes:

• 📂 Complete, production-ready implementations of:

  • Variational Autoencoders (VAE)
  • PatchCore with high-dimensional feature extraction
  • DRAEM for fine-grained anomaly segmentation

• 📓 Comprehensive, didactic Jupyter Notebooks for each method

• 🔧 Track 4: MLOps & Operationalization, including:

  • MLflow for training/inference tracking
  • FastAPI-based model serving
  • Deployment on Render (transferable to Azure/AWS)
  • GitHub Actions for CI/CD

• 🧠 Bonus modules, such as:

  • Loss design and reconstruction trade-offs
  • Dataset visualization and augmentation
  • Deployment checklists and failure cases

💡 This kit helps you move beyond prototypes — toward working, understandable, and extensible anomaly detection pipelines.

👉 🛒 View full product page

---

📂 Dataset: AITEX Fabric Dataset (AFID)

This kit is tailored for the AITEX Fabric Dataset (AFID), which contains high-resolution fabric images with and without labeled defects — ideal for evaluating real-world anomaly detection models.

🔗 Download the Dataset

Due to licensing concerns, the dataset is not included.

Please download it manually from the official AITEX page:
👉 https://www.aitex.es/afid/

⚠️ Licensing Note:
The AITEX dataset may be restricted to non-commercial use. The official site provides no explicit terms.
This demo is intended strictly for educational and evaluation purposes.
Always verify compliance with AITEX licensing before using the dataset in a commercial context.

---

🧰 Dataset Preparation

Once downloaded, follow the steps in notebooks/Dataset Setup.ipynb:

• Extract the archive to the data/ folder
• Organize subdirectories (e.g. Defect, NoDefect, Mask)

---

📚 Notebooks Included in Demo

Notebook	Purpose
01_vae	Full training + inference pipeline with VAE
02_patch_core	Preview of PatchCore logic and structure
03_draem	Preview of DRAEM segmentation logic
04_mlops	Introduction to MLOps + FastAPI inference

---

🚀 Getting Started

Please follow the instructions in Starting the Journey.ipynb.

---

📜 License

See LICENSE.txt for terms.
Source code is flexible. Notebooks are for personal and non-commercial use only.
Commercial use (e.g. resale, redistribution, workshops) requires a separate license.
📩 Contact: [premium-notebooks@grausoft.com]

---

🙌 Acknowledgments

Thanks to AITEX for the dataset.
Thanks to the open-source PyTorch community.