🛑 Research Track Issue
Please read RESEARCH.md before proceeding. Do not submit a Pull Request for this issue without first commenting a brief proposal and receiving approval from the core team.
The Problem:
Evaluating the quality of enhanced PSR images is incredibly difficult because we have no "ground truth" bright images. We currently rely on external libraries for NIQE and BRISQUE, which bottlenecks our evaluation pipeline.
The Research Goal:
We need an ML researcher to write native, GPU-accelerated PyTorch implementations of No-Reference Image Quality Assessment (NR-IQA) metrics.
What you would do:
- Mathematically port the BRISQUE (Blind/Referenceless Image Spatial Quality Evaluator) or NIQE algorithm into pure PyTorch tensor operations.
- Ensure the code supports batched tensors
(B, C, H, W) and executes on the GPU.
- Add this to
src/training/eval.py.
To claim this: Comment below linking to any previous open-source ML work you have done, and confirm your familiarity with PyTorch tensor mathematics.
🛑 Research Track Issue
Please read
RESEARCH.mdbefore proceeding. Do not submit a Pull Request for this issue without first commenting a brief proposal and receiving approval from the core team.The Problem:
Evaluating the quality of enhanced PSR images is incredibly difficult because we have no "ground truth" bright images. We currently rely on external libraries for NIQE and BRISQUE, which bottlenecks our evaluation pipeline.
The Research Goal:
We need an ML researcher to write native, GPU-accelerated PyTorch implementations of No-Reference Image Quality Assessment (NR-IQA) metrics.
What you would do:
(B, C, H, W)and executes on the GPU.src/training/eval.py.To claim this: Comment below linking to any previous open-source ML work you have done, and confirm your familiarity with PyTorch tensor mathematics.