VisionRewardDB-Image is a dataset designed to train VisionReward-Image models. It provides detailed aesthetic annotations across 18 aspects for each image to enhance the assessment of visual quality. The dataset was created by zai-org and last updated on February 24, 2025.
Use Cases
- Train image quality assessment models based on the 18 annotated aesthetic aspects.
- Benchmark visual aesthetic prediction algorithms based on multi-aspect annotations.
- Fine-tune vision-language models for tasks requiring aesthetic judgment based on detailed image attributes.
Strengths
- Each image is annotated across 18 distinct aesthetic aspects.
- Dataset is specifically designed for training VisionReward-Image models.
Limitations
- Column-level documentation is absent; field semantics must be inferred after download.
- Row count is unknown, which may limit suitability assessment.
Provenance
- Source
- zai-org
- Collection Method
- Likely curated for model training; specific gathering method is unknown.
- Time Range
- null
- Freshness
- Last updated 2025-02-24 07:40:04; freshness should be verified.
- Geography
- null