Inspiration

The explosion of deepfakes and manipulated media has made it harder than ever to trust what we see online. From misinformation campaigns to reputational damage, the risks are real. We built DeepVerify to restore digital trust by giving people a reliable way to check authenticity.

What it does

DeepVerify is an AI‑powered verification system that detects manipulated images, videos, and audio. It doesn’t just flag suspicious content — it explains why -using Grad‑CAM visualizations, making AI decisions transparent and trustworthy.

How we built it

Backend: FastAPI in Python for lightweight, scalable APIs. AI Models: CNNs and transformers trained on authentic vs. manipulated datasets. Explainability: Grad‑CAM integrated for visual reasoning. Database: SQL for storing verification logs and metadata.

Challenges we ran into

Efficiently handling large video files without slowing inference. Balancing accuracy with speed for real‑time detection. Making explainability user‑friendly instead of overwhelming. Debugging backend issues like CORS and routing during FastAPI setup.

Accomplishments that we're proud of

Built a working prototype that verifies multiple media types. Achieved strong accuracy while keeping inference times practical. Added explainability features that make AI decisions transparent. Designed a simple, user‑friendly interface for non‑technical users.

What we learned

Human‑centric AI design builds trust when users understand reasoning. Optimizing backend APIs for speed and reliability under pressure. Practical lessons in debugging, database joins, and cloud deployment.

What's next for DEEPVERIFY

Expanding datasets for multilingual and cultural contexts. Adding blockchain‑based verification for immutable authenticity logs. Building browser extensions and social media integrations. Scaling for enterprise use in journalism, cybersecurity, and education.