Deep-learning based IC die identification and die surface defect inspection
Much effort is needed when detecting faults in semiconductors during the manufacturing process. With the rise of artificial intelligence to enhance processes, many would wonder how it is incorporated with semiconductor technologies. However, as a substantial amount of training data is required for t...
Saved in:
| Main Author: | |
|---|---|
| Other Authors: | |
| Format: | Final Year Project |
| Language: | English |
| Published: |
Nanyang Technological University
2023
|
| Subjects: | |
| Online Access: | https://hdl.handle.net/10356/166704 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Institution: | Nanyang Technological University |
| Language: | English |
| Summary: | Much effort is needed when detecting faults in semiconductors during the manufacturing process. With the rise of artificial intelligence to enhance processes, many would wonder how it is incorporated with semiconductor technologies. However, as a substantial amount of training data is required for the detection to be accurate, another concern comes along where a lot of time is required during the process of collecting data and processing these data thereafter.
In this report, a method is proposed to solve the issue of time required during data collection. This enhancement in data collection and processing must not compensate on the accuracy of the model. Tuning of parameters is done to obtain the best combination of values that can offer the best accuracy of the model. In addition, to make benchmarking fair and constant, only one variable is changed during each round of experiment. The outcomes of this study could serve as a basis for creating a roadmap towards advancing the use of convolutional neural networks in deep learning, together with the suggested method for data collection, for detecting semiconductor defects. |
|---|