深度赝品检测

Daksh Baveja,, Yatharth Sharma, Dr. Nagadevi S
{"title":"深度赝品检测","authors":"Daksh Baveja,, Yatharth Sharma, Dr. Nagadevi S","doi":"10.55041/ijsrem36626","DOIUrl":null,"url":null,"abstract":"Abstract—The following paper considers an in-depth study of face detection and classification using a pre-trained VGG16 model with a prime focus on separating real from fake facial images. Face detection is a very fundamental task in computer vision and of key importance in various security- and biometric identification-related applications, social media, and so on, in which the above-mentioned Dortania et al. findings will find their use. The idea is to use transfer learning by tuning an already trained VGG16 that was developed for large-scale image classification to do well in a specific task of face authenticity verification. For this purpose, we constructed a custom dataset with images labeled either ’real’ or ’fake’, sourced from different environments to make it diverse and hence robust. The dataset was then preprocessed by face detection using Haar cascades, resizing, normalization, and augmentation to increase the model’s capacity for generalization. This dataset was trained as well as tested on the modified VGG16 model, where only one fully connected layer at the end was changed to give an output in two classes—one for the real faces and another for the fake ones. Model performance was ascertained through training loss and accuracy in the training phase. For the 30 epochs of training, the model achieved very good training accuracy. Further performance fluctuation analysis at different epochs used detailed plots of the loss and accuracy. Testing validates further that the model is robust, having a high testing accuracy to ensure the model generalizes on unseen data. Our results show the effectiveness of transfer learning using VGG16 in face classification, where accuracy was high for the classification of real and fake faces. Thus, this study not only demonstrates the potential of pre-trained deep models in specialized applications but also shows the proper quality of the dataset and its preprocessing towards the attainment of optimal model performance. This trained model is, therefore, deployable in every real-world application where verification of faces is very important, bringing in a reliable tool for improving security and authenticity in digital relations. Index Terms—deep fake, detection, artificial intelligence, ma- chine learning, digital forensics","PeriodicalId":504501,"journal":{"name":"INTERANTIONAL JOURNAL OF SCIENTIFIC RESEARCH IN ENGINEERING AND MANAGEMENT","volume":"48 4","pages":""},"PeriodicalIF":0.0000,"publicationDate":"2024-07-20","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"1","resultStr":"{\"title\":\"Deep Fake Detection\",\"authors\":\"Daksh Baveja,, Yatharth Sharma, Dr. Nagadevi S\",\"doi\":\"10.55041/ijsrem36626\",\"DOIUrl\":null,\"url\":null,\"abstract\":\"Abstract—The following paper considers an in-depth study of face detection and classification using a pre-trained VGG16 model with a prime focus on separating real from fake facial images. Face detection is a very fundamental task in computer vision and of key importance in various security- and biometric identification-related applications, social media, and so on, in which the above-mentioned Dortania et al. findings will find their use. The idea is to use transfer learning by tuning an already trained VGG16 that was developed for large-scale image classification to do well in a specific task of face authenticity verification. For this purpose, we constructed a custom dataset with images labeled either ’real’ or ’fake’, sourced from different environments to make it diverse and hence robust. The dataset was then preprocessed by face detection using Haar cascades, resizing, normalization, and augmentation to increase the model’s capacity for generalization. This dataset was trained as well as tested on the modified VGG16 model, where only one fully connected layer at the end was changed to give an output in two classes—one for the real faces and another for the fake ones. Model performance was ascertained through training loss and accuracy in the training phase. For the 30 epochs of training, the model achieved very good training accuracy. Further performance fluctuation analysis at different epochs used detailed plots of the loss and accuracy. Testing validates further that the model is robust, having a high testing accuracy to ensure the model generalizes on unseen data. Our results show the effectiveness of transfer learning using VGG16 in face classification, where accuracy was high for the classification of real and fake faces. Thus, this study not only demonstrates the potential of pre-trained deep models in specialized applications but also shows the proper quality of the dataset and its preprocessing towards the attainment of optimal model performance. This trained model is, therefore, deployable in every real-world application where verification of faces is very important, bringing in a reliable tool for improving security and authenticity in digital relations. Index Terms—deep fake, detection, artificial intelligence, ma- chine learning, digital forensics\",\"PeriodicalId\":504501,\"journal\":{\"name\":\"INTERANTIONAL JOURNAL OF SCIENTIFIC RESEARCH IN ENGINEERING AND MANAGEMENT\",\"volume\":\"48 4\",\"pages\":\"\"},\"PeriodicalIF\":0.0000,\"publicationDate\":\"2024-07-20\",\"publicationTypes\":\"Journal Article\",\"fieldsOfStudy\":null,\"isOpenAccess\":false,\"openAccessPdf\":\"\",\"citationCount\":\"1\",\"resultStr\":null,\"platform\":\"Semanticscholar\",\"paperid\":null,\"PeriodicalName\":\"INTERANTIONAL JOURNAL OF SCIENTIFIC RESEARCH IN ENGINEERING AND MANAGEMENT\",\"FirstCategoryId\":\"1085\",\"ListUrlMain\":\"https://doi.org/10.55041/ijsrem36626\",\"RegionNum\":0,\"RegionCategory\":null,\"ArticlePicture\":[],\"TitleCN\":null,\"AbstractTextCN\":null,\"PMCID\":null,\"EPubDate\":\"\",\"PubModel\":\"\",\"JCR\":\"\",\"JCRName\":\"\",\"Score\":null,\"Total\":0}","platform":"Semanticscholar","paperid":null,"PeriodicalName":"INTERANTIONAL JOURNAL OF SCIENTIFIC RESEARCH IN ENGINEERING AND MANAGEMENT","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/10.55041/ijsrem36626","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"","JCRName":"","Score":null,"Total":0}
引用次数: 1

摘要

摘要--本文将利用预先训练好的 VGG16 模型对人脸检测和分类进行深入研究,重点是区分真假人脸图像。人脸检测是计算机视觉中一项非常基本的任务,在各种与安全和生物识别相关的应用、社交媒体等方面具有关键重要性,上述 Dortania 等人的研究成果将在这些应用中得到应用。我们的想法是利用迁移学习,调整已经为大规模图像分类开发的训练有素的 VGG16,使其在特定的人脸真实性验证任务中表现出色。为此,我们构建了一个自定义数据集,其中的图像标注为 "真 "或 "假",这些图像来自不同的环境,因此具有多样性和鲁棒性。然后使用 Haar 级联对数据集进行人脸检测预处理、调整大小、归一化和增强,以提高模型的泛化能力。该数据集在修改后的 VGG16 模型上进行了训练和测试,其中只改变了末端的一个全连接层,以提供两类输出--一类是真人脸,另一类是假人脸。模型的性能通过训练阶段的训练损失和准确率来确定。在 30 个历时的训练中,模型达到了非常高的训练精度。利用损失率和准确率的详细图表,对不同历时的性能波动进行了进一步分析。测试进一步验证了该模型的鲁棒性,它具有很高的测试准确率,可确保模型在未见数据上的泛化。我们的结果表明,在人脸分类中使用 VGG16 进行迁移学习非常有效,对真实和虚假人脸的分类准确率都很高。因此,这项研究不仅证明了预训练深度模型在专业应用中的潜力,还显示了数据集的适当质量及其预处理对获得最佳模型性能的重要性。因此,这种训练有素的模型可以部署到人脸验证非常重要的各种现实应用中,为提高数字关系的安全性和真实性提供了可靠的工具。索引词条-深度伪造、检测、人工智能、机器学习、数字取证
本文章由计算机程序翻译,如有差异,请以英文原文为准。
Deep Fake Detection
Abstract—The following paper considers an in-depth study of face detection and classification using a pre-trained VGG16 model with a prime focus on separating real from fake facial images. Face detection is a very fundamental task in computer vision and of key importance in various security- and biometric identification-related applications, social media, and so on, in which the above-mentioned Dortania et al. findings will find their use. The idea is to use transfer learning by tuning an already trained VGG16 that was developed for large-scale image classification to do well in a specific task of face authenticity verification. For this purpose, we constructed a custom dataset with images labeled either ’real’ or ’fake’, sourced from different environments to make it diverse and hence robust. The dataset was then preprocessed by face detection using Haar cascades, resizing, normalization, and augmentation to increase the model’s capacity for generalization. This dataset was trained as well as tested on the modified VGG16 model, where only one fully connected layer at the end was changed to give an output in two classes—one for the real faces and another for the fake ones. Model performance was ascertained through training loss and accuracy in the training phase. For the 30 epochs of training, the model achieved very good training accuracy. Further performance fluctuation analysis at different epochs used detailed plots of the loss and accuracy. Testing validates further that the model is robust, having a high testing accuracy to ensure the model generalizes on unseen data. Our results show the effectiveness of transfer learning using VGG16 in face classification, where accuracy was high for the classification of real and fake faces. Thus, this study not only demonstrates the potential of pre-trained deep models in specialized applications but also shows the proper quality of the dataset and its preprocessing towards the attainment of optimal model performance. This trained model is, therefore, deployable in every real-world application where verification of faces is very important, bringing in a reliable tool for improving security and authenticity in digital relations. Index Terms—deep fake, detection, artificial intelligence, ma- chine learning, digital forensics
求助全文
通过发布文献求助,成功后即可免费获取论文全文。 去求助
来源期刊
自引率
0.00%
发文量
0
×
引用
GB/T 7714-2015
复制
MLA
复制
APA
复制
导出至
BibTeX EndNote RefMan NoteFirst NoteExpress
×
提示
您的信息不完整,为了账户安全,请先补充。
现在去补充
×
提示
您因"违规操作"
具体请查看互助需知
我知道了
×
提示
确定
请完成安全验证×
copy
已复制链接
快去分享给好友吧!
我知道了
右上角分享
点击右上角分享
0
联系我们:info@booksci.cn Book学术提供免费学术资源搜索服务,方便国内外学者检索中英文文献。致力于提供最便捷和优质的服务体验。 Copyright © 2023 布克学术 All rights reserved.
京ICP备2023020795号-1
ghs 京公网安备 11010802042870号
Book学术文献互助
Book学术文献互助群
群 号:481959085
Book学术官方微信