MISATO: Image Inpainting Dataset for Mitigating Unwanted Object Insertion

The MISATO dataset is proposed in the CVPR 2025 (Highlight) paper "Towards Enhanced Image Inpainting: Mitigating Unwanted Object Insertion and Preserving Color Consistency". Authors from Fudan University created this dataset to address challenges in image inpainting. The dataset page was last updated on 2025-06-24.

Use Cases

• Training image inpainting models to remove objects based on the described focus on mitigating unwanted object insertion.
• Benchmarking inpainting model performance on color consistency preservation tasks.
• Researching methods to improve the realism and coherence of inpainted image regions.

Strengths

• Dataset is associated with a CVPR 2025 Highlight paper, indicating peer-reviewed academic relevance.
• The description explicitly targets two specific challenges in image inpainting: unwanted object insertion and color consistency.

Limitations

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

Provenance

Source

Fudan University

Collection Method

Proposed in a CVPR 2025 paper; specific collection method is not detailed in the provided input.

Freshness

Last updated 2025-06-24 05:41:22; freshness should be verified.