Aesthetic-Train-V2: High-Quality Training Set for Ultra-High-Resolution Image Synthesis

Aesthetic-Train-V2 is a high-quality training set for ultra-high-resolution image generation. The dataset was introduced by author zhang0jhon in the CVPR 2025 paper 'Diffusion-4K: Ultra-High-Resolution Image Synthesis with Latent Diffusion Models'. The dataset was last updated on the Hugging Face platform on June 4, 2025.

Use Cases

• Training diffusion models for ultra-high-resolution image generation based on the dataset's high-quality images.
• Benchmarking image synthesis methods based on the dataset's association with the CVPR 2025 paper.
• Fine-tuning generative models for aesthetic image creation based on the dataset's name and platform tags.

Strengths

• Dataset is associated with a peer-reviewed CVPR 2025 paper, suggesting a research-grade origin.
• Last updated on June 4, 2025, indicating recent maintenance.

Limitations

• Description metadata is limited; actual data quality requires manual inspection after download.
• Column-level documentation is absent; field semantics must be inferred after download.
• Row count, file formats, and license are unknown, which may limit suitability assessment.

Provenance

Source

zhang0jhon on Hugging Face.

Collection Method

Introduced as a training set for the research presented in the CVPR 2025 paper 'Diffusion-4K'.

Time Range

null

Freshness

Last updated 2025-06-04 03:27:33.

Geography

null