NeurIPS-2026-PRISM/PRISM-Dataset

PRISM是一个偏振路面智能感知与测量数据集，包含47,098帧时间同步的多模态数据，结合三色线性偏振、同轴RGB、128通道激光雷达和RTK-GNSS/INS，从车载传感器平台在试验场和开放道路环境中采集，覆盖晴天、阴天、雨天、雾天和雪天条件。该数据集旨在作为一个评估测试平台，用于测量偏振通道是否以及在何种条件下为RGB图像在路面感知中添加判别性信号。PRISM is a polarized pavement intelligent perception and measurement dataset containing 47,098 frames of time-synchronized multimodal data. It integrates four modalities: trichromatic linear polarization, coaxial RGB, 128-channel LiDAR, and RTK-GNSS/INS, and was collected using a vehicle-mounted sensor platform in both proving ground and open-road environments, spanning diverse weather conditions including sunny, cloudy, rainy, foggy and snowy days. This dataset is intended to serve as an evaluation testbed for assessing whether and under what conditions polarization channels can add discriminative signals to RGB images for pavement perception tasks.