The Challenge
Thousands of 19th- and early 20th-century photographs suffered from physical degradation (tears, mold, fading) and required labor-intensive manual restoration. Traditional tools like Photoshop were too slow for bulk processing.
Our Solution
Developed a hybrid pipeline combining GAN-based inpainting (to repair damage) and supervised colorization (trained on period-accurate color references). Incorporated a human-in-the-loop system for quality control.
Technologies Used
Results & Impact
- Restored 45,000 + images at 90%+ fidelity to originals
- Reduced per-image processing time from 2 hours (manual) to <5 minutes
- Enabled public access to previously unusable archival material