Obscura

Inspiration

We noticed how often sensitive information: faces, IDs, license plates, documents, gets shared online without proper redaction. Most tools today are clunky, manual, or require advanced editing skills. We wanted something instant, intuitive, and privacy-first.

What it does

Obscura automatically detects sensitive regions in an image and lets users blur or censor them with a single tap. You stay in control: choose what to hide, what to reveal, and export a clean, privacy-safe version of your image.

How we built it

• React Native app for cross-platform support (mobile + web).
• Expo + Skia for rendering, overlays, and blur effects.
• FastAPI backend powered by computer vision models for object detection.
• Simple UI that balances usability with strong privacy defaults.

Challenges we ran into

• Attempted to integrate diffusion model to generate image to replace censored areas to minimise impact of censoring, but the latency was too high for satisfactory UX.
• Attempted to train a one size fits all model, but the results were not satisfactory

Accomplishments that we're proud of

• Built a working end-to-end pipeline in a short time: pick → analyze → redact → download.
• Designed an interface that feels lightweight but powerful.
• Learned to bridge creative UX with technical efficiency.

What we learned

• Skia is a game-changer for fast image rendering and manipulation in React Native.
• FastAPI as a backend allows for direct model integration and fast-paced development especially for MVPs
• Small UI details (like tap targets and confirmation modals) make a huge difference in user trust. -Privacy isn’t just about algorithms, it’s also about giving users control and confidence.

What's next for Obscura

• Support for video redaction (faces, license plates, screens).
• More redaction styles (pixelation, overlays, custom labels, diffusion generation).
• Batch processing and automation for professionals (journalists, lawyers).

Built With

• expo.io
• fastpi
• python
• react-native
• skia