Overview
What this challenge is about.
You receive 200 paired low-light / well-lit phone photos plus 1,000 unpaired low-light photos. Build a pipeline that combines a learned denoiser (e.g. a small DnCNN-style model or BM3D) with tone mapping and a gentle Retinex-style exposure correction. Quantize or distill the pipeline so it runs under 800 ms on a mid-range Android CPU (Pixel 6a or equivalent). Benchmark against the current naive brightness-boost using a no-reference image quality metric (BRISQUE or NIQE) plus a 12-person blind preference test. Deliver code, benchmark report, and the on-device timing log.
The Brief
What you'll do, and what you'll demonstrate.
Ship a low-light photo restoration pipeline that is faster, more natural, and provably preferred over the current baseline.
Earning criteria — what you'll demonstrate
- Combine classical denoising with learned exposure correction
- Apply on-device optimization techniques (quantization, distillation)
- Evaluate restoration with no-reference quality metrics plus human preference
- Translate research-grade pipelines into mobile-first deployment constraints
Program Fit
Where this fits in your program.
Sharpens the same skills your degree expects you to demonstrate.
Skills
Skills you'll demonstrate.
Each one shows up on your verified credential.
Careers
Roles this prepares you for.
Real titles. Real skill bridges. Pick the one closest to your trajectory.
Computer Vision Engineer
Owning a low-light restoration pipeline end-to-end, with on-device constraints, is the day-one work of a CV engineer at any consumer-AI photo app.
This challenge sharpens
- image-restoration
- denoising
- tone-mapping
Machine Learning Engineer
Quantization, distillation, and mobile deployment are exactly the optimization tasks MLEs ship for consumer apps.
This challenge sharpens
- model-optimization
- benchmarking
- image-restoration
Applied AI Scientist
Pairing no-reference metrics with a blind preference test is the rigorous evaluation an applied AI scientist would defend in a design review.
This challenge sharpens
- benchmarking
- no-reference-quality-metrics
- denoising