Fake Image Detection Using Deep Learning

Created · Updated · completed · computer-vision deep-learning research resnet model-pruning
📑 On this page

Objective

Develop a deep learning model to detect AI-generated fake images with high accuracy, then optimize it for efficient deployment.

Approach

  • CNN architecture: ResNet-50 baseline
  • Two-phase optimization: model pruning (10%, 20%, 30%, 40% ratios) followed by Optuna-driven hyperparameter tuning
  • Dataset: combined GAN (140k Real/Fake Faces) and diffusion (SynthBuster) images, evaluated on the 140k test set
  • Collaboration with Panji Tri Wahyudi and Ida Wahyuni (Institut Teknologi dan Bisnis Asia Malang)

Results

  • Baseline (unpruned) ResNet-50 accuracy: 95.48%
  • All fine-tuned pruned models (up to 40% sparsity) maintained up to 98.87% performance
  • Significant parameter reduction achieved with no loss in accuracy

Outcome

This research was published in ICoBITS: “Implementation of ResNet Architecture for Classification of Real and Synthetic (AI-Generated) Digital Images” (DOI: 10.32664/icobits.v1.58). See the full write-up on the Projects page.

Next Steps

  • Integrate attention mechanisms (Multi-Head Self-Attention) to enhance feature extraction
  • Test robustness against evolving generative techniques