HADataset: High Dynamic Range Images with Per-Pixel Artifact Annotations

HADataset is the first dedicated High Dynamic Range artifact dataset, introduced in a 2025 IEEE Transactions on Image Processing paper. It provides per-pixel annotations for ghosting artifacts in HDR reconstruction, enabling detection-aware optimization strategies. The dataset was created by Xinyueliii and last updated on Hugging Face in February 2026.

Use Cases

• Training artifact detection models based on per-pixel artifact annotations.
• Developing detection-aware HDR reconstruction optimization strategies based on the provided annotations.
• Benchmarking ghosting artifact mitigation methods in HDR imaging.

Strengths

• First dedicated dataset for HDR artifact mitigation, as stated in the description.
• Provides explicit per-pixel artifact annotations, enabling pixel-level analysis.
• Introduced in a peer-reviewed IEEE Transactions on Image Processing paper in 2025.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count and dataset size are unknown, which may limit suitability assessment.

Provenance

Source

Hugging Face user Xinyueliii.

Collection Method

Likely created for the research paper 'Rethinking Artifact Mitigation in HDR Reconstruction: From Detection to Optimization'.

Time Range

null

Freshness

Last updated 2026-02 06 01:44:52; freshness should be verified.

Geography

null