Taxonomy, clustering, and classification of integrated circuit image datasets for deep learning
In the ever-evolving landscape of the semiconductor industry, the integrity of Integrated Circuits (ICs) is critical to technological efficacy and advancement. As such, ensuring the quality of ICs is a high-stakes endeavour where accuracy and precision are paramount. This project introduces a compre...
Saved in:
| Main Author: | |
|---|---|
| Other Authors: | |
| Format: | Final Year Project |
| Language: | English |
| Published: |
Nanyang Technological University
2024
|
| Subjects: | |
| Online Access: | https://hdl.handle.net/10356/177344 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Institution: | Nanyang Technological University |
| Language: | English |
| Summary: | In the ever-evolving landscape of the semiconductor industry, the integrity of Integrated Circuits (ICs) is critical to technological efficacy and advancement. As such, ensuring the quality of ICs is a high-stakes endeavour where accuracy and precision are paramount. This project introduces a comprehensive two-tiered machine learning system, leveraging the robust capabilities of Convolutional Neural Networks (CNNs) to automate the intricate process of defect detection and quality assessment of IC images. Demonstrating an average defect detection accuracy of 96.66% across the Cyclone 3 family of IC types, the system marks a significant advancement over traditional manual inspection methods. Additionally, the Image Quality Assessment (IQA) phase of the system attains an overall accuracy of 78.77%, which is a testament to the system’s prowess in distinguishing between 'low priority defect,' 'minor defect,' and 'okay' classifications. Through the integration of extensive data processing, rigorous model training, and critical validation processes, the system underscores a revolutionary approach to quality control in semiconductor manufacturing, delivering on the promise of enhanced efficiency and reliability. |
|---|