OpenSpaces: Synthetic Visual Question Answering Data with Diverse Image Distribution

OpenSpaces is a synthetic dataset for spatial visual question answering created using VQASynth. It synthesizes data from the first 30,000 rows of the localized narratives split of the cauldron, emphasizing greater diversity in image distribution compared to related datasets. The dataset was authored by remyxai and last updated on October 25, 2024.

Use Cases

• Train visual question answering models based on synthesized spatialVQA data.
• Benchmark model performance on diverse image distributions based on the described emphasis.
• Fine-tune multimodal LLMs using synthetic image-text pairs generated by VQASynth.
• Analyze CLIP embedding similarity for image tagging tasks based on the mentioned labeling method.

Strengths

• Emphasizes greater diversity in image distribution compared to related datasets focusing on warehouse scenes.
• Created using VQASynth, a method for synthesizing spatialVQA data.
• Based on a subset of 30,000 rows from the localized narratives split of the cauldron.

Limitations

• 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.
• Data is synthetic, which may not fully capture the complexity of real-world visual question answering scenarios.

Provenance

Source

huggingface

Collection Method

Synthesized using VQASynth from images in the localized narratives split of the cauldron.

Time Range

null

Freshness

Last updated 2024-10-25 17:21:18; freshness should be verified.

Geography

null