GSI-Bench: Four Computer Vision Evaluation Datasets

GSI-Bench provides four evaluation dataset archives for benchmarking computer vision models. The repository includes datasets named fine_dataset, mesatask_dataset, bathroom_dataset, and robothor_dataset. A small preview split with 16 examples is available for browsing, but the full evaluation data must be downloaded from the provided zip files.

Use Cases

• Benchmarking model performance on visual tasks based on the four provided dataset archives
• Comparing algorithm results against a standardized evaluation suite based on the GSI-Bench code
• Previewing dataset samples for selection before full download based on the 16-example preview split

Strengths

• Includes four distinct evaluation datasets for comprehensive benchmarking
• Provides a 16-example preview split for initial dataset inspection

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

GSI-Bench

Freshness

Last updated 2026-06-03 09:33:24; freshness should be verified