WaferVision: Semiconductor Wafer Defect Detection

📌 Project Overview

WaferVision is an advanced machine learning-powered semiconductor wafer defect detection system. By leveraging computer vision and predictive analytics, this project enhances quality control in semiconductor manufacturing, reducing manual inspection efforts and improving defect identification.

🔑 Key Highlights

✅ XGBoost Classifier trained with 92% accuracy, ensuring precise wafer defect prediction and optimized classification performance. ✅ Image processing techniques (scikit-image) for feature extraction and transformation:

• Radon Transform for defect pattern analysis (skimage.transform.radon)
• Probabilistic Hough Transform for line detection in wafer images (skimage.transform.probabilistic_hough_line)
• Region-based analysis for defect segmentation (skimage.measure) ✅ Statistical analysis (scipy.stats) and interpolation techniques (scipy.interpolate) to refine defect detection and improve model performance. ✅ Streamlit Web Application for an intuitive and interactive user experience, allowing real-time predictions. ✅ Automated Data Preprocessing Pipelines to enhance model accuracy, scalability, and efficiency. ✅ Scalable Architecture designed for industrial applications.

🔧 Tech Stack

• Machine Learning: XGBoost (92% accuracy), Scikit-learn
• Computer Vision: Scikit-image (radon, probabilistic_hough_line, measure), OpenCV
• Statistical & Mathematical Tools: Scipy (interpolate, stats)
• Web Development: Streamlit
• Data Processing: Pandas, NumPy

📂 Project Structure

semicon-wafer-inspect/
│── deployy.py                  # Deployment script
│── requirements.txt            # Dependencies
│── saved_model.json            # Model metadata
│── saved_model_xgb.sav         # Trained XGBoost model
│── savemodle3.sav              # Additional saved model
│── wm-811k-wafermap.ipynb      # Notebook for dataset processing & analysis
│── xgb_model.json              # Trained XGBoost model configuration
│── README.md                   # Project documentation

🚀 Installation & Usage

Step 1: Clone the Repository

git clone https://github.com/ambruhsia/WaferVision.git
cd semicon-wafer-inspect

Step 2: (Optional) Create and Activate a Virtual Environment

python -m venv venv
source venv/bin/activate   # On Windows: venv\Scripts\activate

Step 3: Install Dependencies Ensure you have Python installed (>=3.8), then run:

pip install -r requirements.txt

Running the Model To deploy the wafer defect detection system:

 
python deployy.py

To explore the dataset and analysis:

 
jupyter notebook wm-811k-wafermap.ipynb

🛠 Model Training & Experiment Tracking

• The XGBoost model was trained on the WM-811K wafer dataset, achieving 92% accuracy.
• Preprocessing includes Radon Transform, Hough Transform, and region-based segmentation for feature extraction.
• Statistical techniques were applied to refine classification performance.

📊 Sample Data for Testing

A sample wafer dataset is provided in sample_data.txt(Edge-Ring) for testing the model, or you can get/extract it from WM-811k Wafermap dataset.

🤝 Collaboration

Open to discussions, collaborations, and improvements. Let’s innovate together!

📧 Contact: Reach out for contributions, feedback, or collaboration opportunities.