WeatherReasonSeg

WeatherReasonSeg WeatherReasonSeg is an ECCV 2026 benchmark for weather-aware reasoning segmentation in visual language models. It is designed to evaluate whether a model can still understand a reasoning query and produce an accurate segmentation mask when visual evidence is degraded by fog, rain, snow, or nighttime conditions. This benchmark contains 44,721 image-query pairs and highlights three key aspects: a controllable synthetic subset for severity-aware robustness… See the full description on the dataset page:

Source: Hugging Face Hub (wanwan1111/WeatherReasonSeg). Metadata imported from the dataset’s Hub tags.