Karvat, Mateus (Queen's University), Givigi, Sidney (Queen's University)

Adver-City: Open-Source Multi-Modal Dataset for Collaborative Perception under Adverse Weather Conditions

Scheduled for presentation during the Regular Session "S10c-Cooperative and Connected Autonomous Systems" (WE-LA-T10), Wednesday, November 19, 2025, 16:00−16:20, Cooleangata 4

2025 IEEE 28th International Conference on Intelligent Transportation Systems (ITSC), November 18-21, 2025, Gold Coast, Australia

This information is tentative and subject to change. Compiled on October 19, 2025

Abstract

Adverse weather poses a significant challenge to Autonomous Vehicles (AVs) by degrading sensor performance and perception reliability. While Collaborative Perception (CP) helps mitigate this, existing CP datasets lack coverage of adverse weather conditions. To address this gap, we introduce Adver-City, the first open-source synthetic CP dataset focused on adverse weather. Simulated in CARLA with OpenCDA, it contains 24,000+ frames, 890,000+ annotations, and 110 unique scenarios across six weather conditions: clear, soft rain, heavy rain, fog, foggy heavy rain, and, for the first time in a synthetic CP dataset, glare. It has six object categories including pedestrians and cyclists, and uses data from vehicles and roadside units featuring LiDARs, RGB and semantic segmentation cameras, GNSS, and IMUs. Scenarios are based on real crash reports, depicting the most relevant road configurations for adverse weather and poor visibility. They include both dense and sparse scenes. These characteristics allow for novel testing conditions of CP models. Benchmark results confirm the dataset's difficulty. CoBEVT achieved 58.30/52.44/38.90 (AP@30/50/70). Dataset, code and documentation are available at https://labs.cs.queensu.ca/quarrg/datasets/adver-city/.