Synthetic Visual Genome (SVG) datasets are designed for training Vision-Language Models on scene graph understanding and dense visual relationships. The datasets were created by author jamepark3922 and were last updated on June 11, 2025. They are hosted on the Hugging Face platform.
Use Cases
- Training scene graph generation models based on the described focus on visual relationships.
- Benchmarking dense visual relationship detection models based on the dataset's stated purpose.
- Pre-training Vision-Language Models (VLMs) on synthetic structured visual data as indicated by the description.
Strengths
- Dataset is explicitly designed for two complementary tasks: scene graph understanding and dense visual relationships.
- Associated resources include training code, a model checkpoint (ROBIN-3b), and a research paper, suggesting a research-backed foundation.
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
- huggingface
- Collection Method
- Synthetic generation, as indicated by the dataset title and description.
- Freshness
- Last updated 2025-06-11 05:30:10; freshness should be verified.