TIGER-Lab's OmniEdit Filtered 1.2M dataset, last updated December 2024, is designed for training a general-purpose image editing model. The dataset was created by filtering data using large multimodal models like GPT-4o for quality assessment. It provides supervision for seven distinct image editing tasks.
Use Cases
- Train a unified image editor based on supervision from seven specialist models mentioned in the description
- Benchmark filtering techniques for training data using large multimodal model scores as described
- Develop models capable of handling image editing at any aspect ratio as referenced in the description
Strengths
- Dataset is associated with a published research paper and project website.
- Data quality was improved using importance sampling based on scores from large multimodal models like GPT-4o.
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 is unknown, which may limit suitability assessment.
Provenance
- Source
- TIGER-Lab
- Collection Method
- Filtered using supervision from specialist models and importance sampling with large multimodal model scores.
- Time Range
- null
- Freshness
- Last updated 2024-12-06 02:57:59; freshness should be verified.
- Geography
- null