Please use this identifier to cite or link to this item:
http://hdl.handle.net/10603/541442
Title: | An Intelligent Approach to Detect and Classify Facial Image Forgery |
Researcher: | Sheth Kinjal Ravi |
Guide(s): | Dr. Vishal S Vora |
Keywords: | Adam Engineering Engineering and Technology Engineering Electrical and Electronic Fine tuning Optimizer Retouching RMSprop Transfer Learning VGG16 |
University: | Atmiya University |
Completed Date: | 2023 |
Abstract: | the era of digital image manipulation and the pervasive presence of social media, the verification newlineof facial image authenticity has become an essential concern. Facial retouching in supporting newlinedocuments can have adverse effects, undermining the credibility and authenticity of the information newlinepresented. In official identification documents, such as passports or driverand#39;s licenses, retouching can newlinehinder accurate identification and security measures, potentially leading to identity fraud and security newlinerisks. newline newlineThis study presents a comprehensive investigation into the classification of retouched face images newlineusing a fine-tuned pre-trained VGG16 andamp; ResNet50 model with ImageNet weight. We explore the newlineimpact of different train-test split strategies on the performance of the model and also evaluate the newlineeffectiveness of two distinct optimizers namely Adam and RMSprop. The model generalizability has newlinebeen checked over two standard datasets ND-IIITD retouched faces and MDRF (Multi Demography newlineRetouched Faces- Caucasian samples). newline newlineThe experiment results indicate that the ResNet50 model, fine-tuned with the RMSprop optimizer, newlineattains a maximum accuracy of 98.52% for ND-IIITD and an impressive 99.17% for MDRF newline(Caucasian). In addition, an examination of various train-test split ratios over these datasets reveals newlinethe 80%-20% split ratio as the optimal choice for the approach. Moreover, the experiments show that newlinethis method effectively performs on both balanced and imbalanced datasets, emphasizing its newlinerobustness and adaptability. newline newlineIn conclusion, the intelligent approach leverages transfer learning and model selection offers a robust newlinesolution for the automated detection and classification of facial retouching. This contribution not newlineonly enhances image authenticity and trustworthiness in the digital age but also emphasizes the newlineimportance of considering various factors, such as model selection and dataset characteristics and newlinehyperparameters in achieving optimal results in this field. newline |
Pagination: | A4 |
URI: | http://hdl.handle.net/10603/541442 |
Appears in Departments: | Electronics & Communication Engineering |
Files in This Item:
File | Description | Size | Format | |
---|---|---|---|---|
01_title.pdf | Attached File | 170.78 kB | Adobe PDF | View/Open |
02_prelim pages.pdf | 2.62 MB | Adobe PDF | View/Open | |
03_content.pdf | 199.84 kB | Adobe PDF | View/Open | |
04_abstract.pdf | 185.46 kB | Adobe PDF | View/Open | |
05_chapter1.pdf | 331.97 kB | Adobe PDF | View/Open | |
06_chapter 2.pdf | 576.24 kB | Adobe PDF | View/Open | |
07_chapter 3.pdf | 447.04 kB | Adobe PDF | View/Open | |
08_chapter 4.pdf | 822.36 kB | Adobe PDF | View/Open | |
09_chapter 5.pdf | 1.56 MB | Adobe PDF | View/Open | |
10_chapter 6.pdf | 222.17 kB | Adobe PDF | View/Open | |
80_recommendation.pdf | 455.64 kB | Adobe PDF | View/Open |
Items in Shodhganga are licensed under Creative Commons Licence Attribution-NonCommercial-ShareAlike 4.0 International (CC BY-NC-SA 4.0).
Altmetric Badge: