GitHub Repositories
Open source projects and research code from Reality Lab
SIDL Benchmark
Smartphone Dirty Lens Dataset
🎉 AAAI25 Accept!
Realitylab-site
Main Website Repository
Computer Vision
Deep Learning & Vision
Robotics & AI
Autonomous Systems
Machine Learning
AI Algorithms & Models
Research Papers
Publication Code
Dev Tools
Utilities & Libraries
SIDL Benchmark
Smartphone Images with Dirty Lenses Dataset
Input (Contaminated)
Target (Restored)
Description: A novel dataset designed to restore images captured through contaminated smartphone lenses with diverse real-world contaminants.
Technologies: Computer Vision, Image Restoration, Machine Learning, Dataset
Key Features:
- 300 static scenes with 1,588 degraded-clean image pairs
- Full ProRAW resolution (4032 × 3024, 12-bit DNG)
- 4 contaminant types: fingerprints, dust, scratches, water drops
- Difficulty levels: Easy, Medium, Hard
- Comprehensive evaluation framework
