A Web Application for Fake Face Detection

We are aiming to develop a web application for Real and Fake Face Detection utilizing either Flask or Streamlit, based on TensorFlow or PyTorch.

Dataset Collection: Initially, we plan to reuse an open-source profile image dataset previously compiled from Kaggle and Twitter profiles (Seonghyeon, Jan 2019).
Model Choice: Our focus will be on leveraging a Convolutional Neural Network (CNN) for image classification. Specifically, we intend to start with a pretrained model, such as EfficientNet, and then fine-tune it.
Training: We will train our model using the Real and Fake Face Image dataset, ensuring a balanced representation of real and fake images.
Evaluation: The model's accuracy will be assessed and enhanced through fine-tuning and optimization. Benchmarking efforts may also be conducted to evaluate performance further.
Deployment: Once the model is adequately fine-tuned, we will employ Streamlit, a lightweight framework, to create a user-friendly application. This app will allow users to upload images to determine whether they are real or AI-generated. We might use GitHub Page as our web server.

Name		Name	Last commit message	Last commit date
Latest commit History 35 Commits
Report		Report
assets/data		assets/data
23WS_CV3D_Fake_Face_Detection_JiufengV1.ipynb		23WS_CV3D_Fake_Face_Detection_JiufengV1.ipynb
23WS_CV3D_Fake_Face_Detection_JiufengV2.ipynb		23WS_CV3D_Fake_Face_Detection_JiufengV2.ipynb
README.md		README.md
main-WIP-P-v1.ipynb		main-WIP-P-v1.ipynb
main-WIP-P-v2.ipynb		main-WIP-P-v2.ipynb
main-WIP-P-v3.ipynb		main-WIP-P-v3.ipynb
main-WIP-P-v4.ipynb		main-WIP-P-v4.ipynb
main-WIP-P-v5.ipynb		main-WIP-P-v5.ipynb
model_final.ipynb		model_final.ipynb
model_improvedcnn_d2_complete.pth		model_improvedcnn_d2_complete.pth

Provide feedback