Shuai1995 built a dataset of 66,000 images for image aesthetics assessment (IAA), released on Hugging Face in March 2023. The dataset is theme-oriented, covering 47 popular themes, with images selected by hand. Each theme has its own aesthetic criteria, and images are annotated independently.
Use Cases
- Train image aesthetics assessment models based on theme-specific criteria.
- Benchmark theme classification models based on 47 curated categories.
- Study the relationship between image themes and perceived aesthetic quality.
- Develop multi-task models for joint theme and aesthetics prediction.
Strengths
- Contains 66,000 images, providing a substantial scale for model training.
- Covers 47 distinct themes, offering a structured, multi-category framework.
- Images were hand-selected based on theme, suggesting a degree of curation.
Limitations
- Column-level documentation is absent; field semantics must be inferred after download.
- Row count is unknown, which may limit suitability assessment.
- Last updated 2023-03-30 01:24:17; freshness should be verified.
Provenance
- Source
- Hugging Face, author Shuai1995.
- Collection Method
- Images were hand-selected based on theme and annotated.
- Time Range
- null
- Freshness
- Last updated 2023-03-30 01:24:17.
- Geography
- null