WeatherReasonSeg: A Benchmark for Weather-Aware Reasoning Segmentation

44,721 image-query pairs form a benchmark for evaluating visual language models under degraded weather conditions. WeatherReasonSeg was created by wanwan1111 and announced for ECCV 2026. It tests model robustness to fog, rain, snow, and nighttime visual degradation.

Use Cases

• Benchmarking model robustness to weather degradation based on the synthetic subset with controllable severity.
• Evaluating reasoning segmentation accuracy based on the image-query pairs under adverse conditions.
• Training models for scene understanding in poor weather based on the described fog, rain, snow, and nighttime scenarios.

Strengths

• Contains 44,721 image-query pairs for evaluation.
• Includes a controllable synthetic subset for severity-aware robustness testing.
• Focuses on four specific weather degradation conditions: fog, rain, snow, and nighttime.

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

Likely contains synthetic and/or curated image-query pairs for benchmark purposes.

Freshness

Last updated 2026-06-22 11:48:31; freshness should be verified.