Artificial Intelligence Generated Image Detection
V ROSHAN KUMAR*1, SYPU SUSMITHA2, VARI NAGA PUJITHA 3, PASAM GOPI CHAND4, YARRAREDDI LAKSHMI SURENDRA 5
1Assistant Professor, Department of CSE(AIML), Bapatla Engineering College, Bapatla 522101, AP,India
2Student, Department of CSE(AIML),Bapatla Engineering College, Bapatla 522101,AP, India
3Student, Department of CSE(AIML), Bapatla Engineering College, Bapatla 522101, AP, India
4Student, Department of CSE(AIML), Bapatla Engineering College, Bapatla 522101, AP, India
5Student, Department of CSE(AIML), Bapatla Engineering College, Bapatla 522101, AP, India
Abstract—This project introduces an intelligent system for detecting whether an image is real or AI-generated by combining deep learning with image feature analysis. The approach is based on a Convolutional Neural Network (CNN) model that predicts image authenticity, supported by additional handcrafted features such as noise score, edge density, brightness, contrast, and color variance. These features help capture subtle visual differences between natural images and synthetic outputs.
A key contribution of this system is the integration of explainable AI, where the model not only provides a classification result but also generates clear, human-readable reasons for its decision. This improves transparency and user trust in the system. The framework also includes an automated report generation module that produces detailed PDF reports containing the prediction, confidence level, feature analysis, and explanation.
The implementation utilizes TensorFlow for model inference, OpenCV for feature extraction, and Gradio to create an interactive interface supporting batch image uploads. This combined approach enhances detection accuracy while ensuring usability and interpretability. The system can be applied in digital forensics, content verification, and detection of AI-generated media, contributing to more reliable and trustworthy image analysis solutions
Keywords:Image detection, Convolutional Neural Networks (CNNs), Local features, Share weights, Pooling mechanisms, Innovative approach, ShortCut3-ResNet, Residual Network (ResNet), Feature extraction, PDF Report Generation.
