589 multi-exposure image sequences containing 4,413 high-resolution images paired with high-quality reference images generated via fusion. The dataset covers diverse scenes including indoor, outdoor, and architectural environments to train models for single-image contrast enhancement.
Use Cases
- Train deep learning models for single-image contrast enhancement using the multi-exposure sequences as input and fused images as ground truth
- Evaluate image quality assessment (IQA) metrics on contrast-enhanced results compared to the provided reference images
- Develop exposure correction algorithms by mapping under-exposed or over-exposed frames to the high-quality target
Strengths
- 589 multi-exposure image sequences with varying exposure levels
- 4,413 total images used for training and evaluation of contrast enhancement algorithms
- High-quality reference images generated through multi-exposure fusion (MEF) and stack-based HDR methods
- Diverse scene categories including indoor, outdoor, and low-light environments