patil.pptx.deep fake image on ppt slide share

Image

1. AUTHOR : KAVYA SREE PATIL CO-AUTHORS : ANUSHA THANDRAPATI ASHWINI PURIMETLA JNTUA COLLEGE OF ENGINEERING KALIKIRI COMPUTER SCIENCE AND ENGINEERING DEEPFAKE IMAGE DETECTION

2. OUTLINE INTRODUCTION OBJECTIVE OF THE WORK PROPOSED METHODOLOGY RESULTS CONCLUSION REFERENCES

3. INTRODUCTION The recent growth of technology in computer-generated editing programs has made synthesizing and modifying media content easier than ever. The potential for misinformation spread has exploded, especially with the phenomenon known as Deepfake . Deepfake is a technology that uses deep learning to create fake images, alter existing images.

4. OBJECTIVE OF THE WORK Develop a robust deepfake image detection system using Convolutional Neural Network (CNN) architecture Enhance the detection capability of the model by training it on a diverse dataset comprising both genuine and synthetic images, encompassing various facial expressions, lighting conditions, and backgrounds. Implement transfer learning techniques to leverage the pre-trained VGG model,ResNet50 model,Inception_V3, on ImageNet, optimizing the training process and improving the model's performance in detecting deepfake images.

5. PROPOSED METHODOLOGY 1 Data Collection Gather a diverse dataset of real & deepfake images . 2 Preprocessing DIMENSIONALLITY REDUCTION NORMALIZATION ENHANCING THE PERFORMANE 3 Training and Evaluation Choose a pre-trained convolutional neural network model such as VGG, ResNet50,or Inception, which has been trained on a large dataset like ImageNet.

6. VGG_19: ACCURACY:86% RESULTS MODEL PERFORMANCES METRICS:

7. INCEPTION_V3: ACCURACY :82.6%

8. ResNet50 : ACCURACY :72.6%

9. OUTPUT:

10. CONLUSION In conclusion, our study on deep fake detection has demonstrated promising results, particularly when leveraging the VGG-19 architecture. Through rigorous experimentation and evaluation, we found that VGG-19 outperformed ResNet50 and InceptionV3 models in accurately detecting deep fake images. With an impressive accuracy rate of 86%, the VGG-19 model showcased its effectiveness in discerning between genuine and manipulated images. This superior performance underscores the significance of choosing the appropriate convolutional neural network (CNN) architecture for deep fake detection tasks.

11. REFERENCE https://www.tensorflow.org/api_docs/python/tf/keras/applications/vgg19 https://ieeexplore.ieee.org/document/9776410 Kumari, R., Ekblad, A. (2021). Amba: Attention-based multimodal factorized bilinear pooling for multimodal https://doi.org/10.1016/j.eswa.2021.115412 Fake_Image_Detection_Using_Machine_Learn[1].pdf

12. ANY QUESTIONS