mpahlevi64/LowLevelBanana
🍌 Explore LowLevelBanana, a comprehensive evaluation tool for low-level vision tasks across 14 tasks and 40 datasets in AI research.
What's novel
🍌 Explore LowLevelBanana, a comprehensive evaluation tool for low-level vision tasks across 14 tasks and 40 datasets in AI research.
Code Analysis
15 files read · 4 roundsA personal collection of ~14 standalone Python/MATLAB scripts that compute image quality metrics (PSNR, SSIM, LPIPS, NIQE, BRISQUE, NIMA, UCIQE, UIQM, VIFP, NMI, QY, QCB, DeltaE) for various low-level vision tasks, each with hardcoded file paths and no unified interface.
Strengths
The multi-focus fusion and infrared-visible fusion scripts implement genuinely non-trivial quality metrics (QY, QCB, VIFP, SCD) from scratch using PyTorch, and the underwater enhancement script correctly implements UCIQE/UIQM sub-metrics. The NIQE implementation is a faithful port of the original algorithm.
Weaknesses
Massive code duplication (identical files for denoising/low-light, duplicated basicsr package, repeated utility functions), zero tests, hardcoded paths throughout, no CLI or configuration system, and the README describes a GUI application with 40 datasets that does not exist in the code.
Score Breakdown
Signal breakdown
Innovation
Craft
Traction
Scope
Evidence
Commits
22
Contributors
2
Files
122
Active weeks
4
Repository
Language
Python
Stars
1
Forks
0
License
Apache-2.0