TrueSight: A Multi-Modal Framework for AI Image Authenticity Detection

This paper presented TrueSight, an integrated AI vision system combining deep learning and hybrid object detection. The system achieves high accuracy and provides explainable results. Future work includes real-time optimization and video analysis

[1] A. Radford et al., “Learning Transferable Visual Models From Natural Language Supervision (CLIP),” 2021.  https://arxiv.org/abs/2103.00020    [2] M. Tan and Q. Le, “EfficientNet: Rethinking Model Scaling for Convolutional Neural Networks,” 2019.  https://arxiv.org/abs/1905.11946    [3] I. Goodfellow et al., “Generative Adversarial Networks,” 2014.  https://arxiv.org/abs/1406.2661    [4] J. Ho et al., “Denoising Diffusion Probabilistic Models,” 2020.  https://arxiv.org/abs/2006.11239    [5] A. Dosovitskiy et al., “An Image is Worth 16x16 Words: Vision Transformer,” 2020.  https://arxiv.org/abs/2010.11929    [6] D. G. Lowe, “Distinctive Image Features from Scale-Invariant Keypoints (SIFT),” 2004.  https://www.cs.ubc.ca/~lowe/papers/ijcv04.pdf    [7] P. Viola and M. Jones, “Rapid Object Detection using a Boosted Cascade of Simple Features,” 2001.  https://ieeexplore.ieee.org/document/990517    [8] F. Chollet, “Xception: Deep Learning with Depthwise Separable Convolutions,” 2017.  https://arxiv.org/abs/1610.02357    [9] Bayar and Stamm, “A Deep Learning Approach to Universal Image Manipulation Detection,” 2016.  https://arxiv.org/abs/1609.07093    [10] Cozzolino et al., “Noiseprint: CNN-based Camera Model Fingerprint,” 2018.  https://arxiv.org/abs/1811.07853