We use MixedWM38, the mixed-type wafer defect pattern dataset for wafer defect pattern regcognition with visual transformers.
-
Updated
Oct 1, 2023 - Jupyter Notebook
We use MixedWM38, the mixed-type wafer defect pattern dataset for wafer defect pattern regcognition with visual transformers.
A taxonomy of defects with a benchmarking script that validates which of them can be spot by which static analyzers
Diameter measurement, cable defect detection and classification.
A user-friendly web application for detecting defects on painted surfaces using live webcam input and image uploads.
An impactful project SEM -i-CON which does Defect Classification for Semiconductor Wafer/Die Images from Specialized Electron microscope (SEM) by developing an edge AI
This repository contains implementation of ResNet for surface defect classification, with detailed analysis of results.
With only 4 real SEM images of a defect class, do generated images help the classifier? A pre-registered experiment on Carinthia-S with NVIDIA Cosmos AnomalyGen.
This project is for defect classification on multi-light source
Official implementation of DiffWaferNet, a wafermap defect classification framework using CNN backbone with differential attention and CVAE-based synthetic data generation. [PeerJ Computer Science]
Wafer-map yield analysis on WM-811K: yield-loss split, lot excursions, defect-pattern classification (random forest / CNN), spatial and tool-commonality tests
Steel-plate defect classification on real UCI data (1,941 plates): frozen holdout macro-F1 0.815, human-review triage, Power BI layer, ONNX export and a dbt data contract.
🚀 Official code for IEEE ECCE 2025 paper: "Advancing Semiconductor Fabrication: A CNN-Based Wafer Defect Detection with XAI Insights". ResNet50 classifier (94.08% accuracy) with Explainable AI (Grad-CAM) on the WM-811K benchmark.
ResNet-18 wafer defect classifier on WM-811K: focal loss, CBAM attention, temperature-scaled calibration, Grad-CAM++. Macro-F1 0.92
Mango Classification Competition by AICUP 2020
Wafer-map defect classification on WM-811K (LightGBM vs CNN, lot-grouped evaluation) served through a FastAPI + React yield-analytics app.
An end-to-end, production-ready MLOps ecosystem designed for automated Metal surface inspection.
To associate your repository with the defect-classification topic, visit your repo's landing page and select "manage topics."