WM811k-Silicon-Wafer-Map-Dataset-and-Defect-Detection-Using-CNNs-with-Grad-CAM

This repository provides the WM811k silicon wafer map dataset subset, MATLAB implementation, and Jupyter Notebook(code in MATLAB software) for defect detection in semiconductor wafers using Convolutional Neural Networks (CNNs). The research is based on the paper:

Enhancing Defect Recognition: Convolutional Neural Networks for Silicon Wafer Map Analysis
Published in the 2024 3rd International Conference on Advancement in Electrical and Electronic Engineering (ICAEEE).

---

Features

1. Dataset:

   • A curated subset of the WM811k silicon wafer map dataset with 902 images categorized into 9 defect classes:

     • Center
     • Donut
     • Edge Local
     • Edge Ring
     • Local
     • Near Full
     • None
     • Random
     • Scratch

   • Images are resized to 32x32 pixels for efficient processing.

2. MATLAB Code (grad.m):

   • Implements a CNN architecture with five convolutional layers, batch normalization, ReLU activation, and Grad-CAM visualization. MATLAB Implementation¶

     Run Code in MATLAB to: >Train the CNN

      >valuate its performance
     
     > Visualize Grad-CAM results
     

3. Jupyter Notebook (defect-detection-using-cnns-and-gradcam-vis.ipynb):

   • Provides an interactive Python-based implementation of defect detection and Grad-CAM visualization.
   • Useful for researchers who prefer Python-based workflows.

4. Research Paper:

   • Enhancing_Defect_Recognition_Convolutional_Neural_Networks_for_Silicon_Wafer_Map_Analysis.pdf

---

Repository Structure

WM811k-Defect-Detection/ ├── Dataset/ │ ├── Center/ │ ├── Donut/ │ ├── Edge_Local/ │ ├── Edge_Ring/ │ ├── Local/ │ ├── Near_Full/ │ ├── None/ │ ├── Random/ │ └── Scratch/ ├── grad.m ├── defect-detection-using-cnns-and-gradcam-vis.ipynb ├── README.md └── Paper/ └── Enhancing_Defect_Recognition_CNN.pdf