STVQA-7K: Spatial Visual Question Answering Dataset with 7,587 Samples

STVQA-7K is a high-quality spatial visual question answering dataset comprising 7,587 samples. It was created by hunarbatra and last updated on 2026-01-29. The dataset is fully grounded in human-annotated scene graphs from Visual Genome and is designed for training and evaluating spatial reasoning capabilities in multimodal large language models.

Use Cases

• Training multimodal LLMs for spatial reasoning based on the dataset's 9 distinct spatial reasoning categories.
• Benchmarking model performance on visual question answering tasks grounded in scene graphs.
• Analyzing the relationship between visual scene structure and language-based spatial queries.
• Developing specialized models for interpreting spatial relationships in images.

Strengths

• Contains 7,587 high-quality, fully grounded samples.
• Based on human-annotated scene graphs from the established Visual Genome dataset.
• Designed to cover 9 distinct spatial reasoning categories.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is known, but other details like file formats and license are unknown.
• Last updated 2026-01-29 10:38:40; freshness should be verified.

Provenance

Source

Visual Genome scene graphs, via huggingface user hunarbatra.

Collection Method

Human-annotated scene graphs.

Time Range

null

Freshness

Last updated 2026-01-29 10:38:40.

Geography

null