Scribd
Upload
7 views3 pages

Project Report Final 1 3

This project report presents the use of Siamese Networks, combined with ResNet18, for detecting manufacturing defects to improve classification accuracy. The study demonstrates that this approach achieves an accuracy of 94.4%, outperforming traditional CNN methods, particularly in identifying subtle defects. The findings suggest that the Siamese Network offers significant advantages for quality control in manufacturing through effective pairwise image comparisons.

Uploaded by

raji7balag
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
7 views3 pages

Project Report Final 1 3

This project report presents the use of Siamese Networks, combined with ResNet18, for detecting manufacturing defects to improve classification accuracy. The study demonstrates that this approach achieves an accuracy of 94.4%, outperforming traditional CNN methods, particularly in identifying subtle defects. The findings suggest that the Siamese Network offers significant advantages for quality control in manufacturing through effective pairwise image comparisons.

Uploaded by

raji7balag
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 3

DETECTION OF MANUFACTURING DEFECTS USING

SIAMESE NETWORK

A PROJECT REPORT

Submitted by

RAJA MARIAPPAN T
(2023178056)

A report for the project


submitted to the Faculty of

INFORMATION AND COMMUNICATION ENGINEERING

in partial fulfillment
for the award of the degree
of

MASTER OF COMPUTER APPLICATIONS

DEPARTMENT OF INFORMATION SCIENCE AND TECHNOLOGY


COLLEGE OF ENGINEERING, GUINDY
ANNA UNIVERSITY
CHENNAI 600 025
APRIL 2025
ii

ANNA UNIVERSITY
CHENNAI - 600 025
BONAFIDE CERTIFICATE

Certified that this project report titled ”Detection of Manufacturing


defects using Siamese Network” is the bonafide work of Raja Mariappn
(2023178056) who carried out project work under my supervision. Certified
further that to the best of my knowledge and belief, the work reported herein
does not form part of any other thesis or dissertation on the basis of which a
degree or an award was conferred on an earlier occasion on this or any other
candidate.

PLACE: Dr. M. DEIVAMANI


DATE: ASSISTANT PROFESSOR
PROJECT GUIDE
DEPARTMENT OF IST, CEG
ANNA UNIVERSITY
CHENNAI 600025

COUNTERSIGNED

Dr.S. SWAMYNATHAN
HEAD OF THE DEPARTMENT
DEPARTMENT OF INFORMATION SCIENCE AND TECHNOLOGY
COLLEGE OF ENGINEERING, GUINDY
ANNA UNIVERSITY
CHENNAI 600025
iii

ABSTRACT

Automated defect detection plays a crucial role in modern


manufacturing, ensuring high-quality production standards with minimal human
intervention. Traditional image classification models often struggle with
variability in defect appearances, leading to inconsistent detection rates.
This study explores the effectiveness of Siamese Networks, integrated with
ResNet18 as a feature extractor, to enhance defect classification accuracy.
Unlike conventional models that rely on direct classification, the Siamese
architecture leverages pairwise image comparisons, learning to differentiate
between defective and non-defective items based on structural similarities. This
approach enables the model to generalize better across diverse defect types,
making it more robust for real-world applications.

The proposed framework extracts deep features from input


image pairs, computes similarity scores, and classifies defects with higher
precision. Experimental evaluations on our test dataset demonstrate promising
improvements, achieving an overall accuracy of 94.4% compared to 87.6%
from conventional CNN approaches. Observation of the model’s performance
showed that the Siamese Network was particularly effective at identifying
subtle defects that were missed by traditional methods. Analysis of the results
revealed that the model performed best when provided with clear reference
images of non-defective products. Additionally, the similarity threshold of
0.7 was found to provide optimal balance between false positives and false
negatives. Our findings confirm that the Siamese Network approach offers
significant advantages for manufacturing defect detection through simple image
pair comparison, making it an accessible yet powerful tool for quality control
applications.

You might also like