RefCOCO-Degraded: Images with Fog, Smoke, and Thermal Degradation for VLM Robustness

RefCOCO-Degraded is a benchmark dataset for evaluating the robustness of vision-language models. Images are modified with simulated degradation effects such as fog, smoke, and thermal noise. The dataset's author, organization, and specific scale are not provided in the input metadata.

Use Cases

• Benchmarking model robustness to fog and smoke based on the described image degradation types
• Evaluating visual grounding performance under thermal degradation based on the described image degradation types
• Training models to be invariant to specific environmental noise based on the described degradation effects

Strengths

• Focuses on specific, named degradation types (fog, smoke, thermal) for controlled robustness testing
• Designed for a clear evaluation purpose: benchmarking vision-language model robustness

Limitations

• Row count is unknown, which may limit suitability assessment
• Column-level documentation is absent; field semantics must be inferred after download
• Last update date is unknown; freshness unverified

Provenance

Collection Method

Images likely contain simulated degradation effects, as described.