Deepfake Detection IEEE SPS Cup 2025
Built deep learning pipelines for facial manipulation detection using transfer learning under IEEE competition constraints.
This project involved developing detection pipelines for deepfakes and facial manipulations as part of the IEEE SPS Cup 2025 competition. The challenge required identifying synthesized or manipulated faces in images and videos, a critical problem as synthetic media becomes increasingly convincing.
We implemented transfer learning approaches using pre-trained models fine-tuned on competition datasets, experimented with ensemble methods, and optimized inference pipelines for real-time detection. The work highlighted the arms race between generative models and detection systems, and the importance of developing robust defenses as manipulation techniques become more sophisticated.
In stealth due to licensing issues :(