11,125 high-resolution smartphone images across 11 scene categories, each annotated with subjective Mean Opinion Scores (MOS). The dataset includes five image attribute labels—brightness, color, contrast, sharpness, and noise—captured using 66 different mobile devices.
Use Cases
- Train blind image quality assessment (BIQA) algorithms using the MOS column as the target variable
- Perform multi-attribute regression to predict sharpness and noise levels from raw image pixels
- Correlate smartphone hardware specifications from EXIF data with perceived image quality scores
Strengths
- 11,125 high-resolution images captured by 66 distinct smartphone models
- Subjective quality ratings provided as Mean Opinion Scores (MOS)
- Fine-grained labels for 5 attributes: brightness, color, contrast, sharpness, and noise
- Includes EXIF metadata such as ISO, aperture, and exposure time for hardware-level analysis