In a world where artificial intelligence can generate virtually anything, realistic synthetic videos have become a threat to trust in digital spaces. This project uses spatiotemporal content features and style based facial attributes to create a video deepfake detection system that can identify these manipulations. The model achieves 88% and 91% accuracy on Celeb-DF and FaceForensics+ datasets respectively, this work proves a new alternative for detecting deepfakes with high efficiency.
The inefficiency and personnel demand for legal documentation analysis & review has led to growing interest in artificial intelligence (AI) and large language models (LLMs) as tools for automating analysis and providing and decision support for valuable legal experts. This project includes an examination of recent work on AI legal document processing, focusing on feasibility, reliability, effectiveness and ethical considerations. This project uses a large language model to analyze documents, evaluate their contents, and perform decision pathing across an attorney-written checklist.