Back to Ecosystem Pulse
EvaluationExperimental
Adver-City
by QUARRG
Synthetic multi-agent collaborative perception dataset focused on adverse weather
Python
Updated Apr 29, 2025
Share:
Overview
Provides the first open-source synthetic dataset for collaborative perception under adverse weather. Combines multi-modal sensor streams (camera, lidar) and multi-agent viewpoints to simulate V2V/V2I cooperative perception failures and degradations. Includes annotated scenarios designed to stress sensor fusion and inter-agent message exchange in rain, fog, and low-visibility conditions.
Key Benefits
As agents rely on exchanged perception data, adverse weather exposes brittle failure modes that single-vehicle benchmarks miss. Until now, there were few public datasets focused on cooperative perception under challenging conditions, making it hard to evaluate agent-to-agent reliability and track records. This dataset lets researchers and QA teams benchmark A2A evaluation patterns, probe agent failure modes, and develop trust signals for multi-agent perception.
When to Use
Researchers and engineers testing cooperative perception, sensor-fusion algorithms, and A2A evaluation under degraded environmental conditions.
Applications
- Benchmark cooperative perception algorithms under rain, fog, and low visibility
- Evaluate agent-to-agent message loss and fusion strategies for A2A evaluation
- Reproduce multi-agent system failures to develop agent reliability and trust signals
Topics
automated-driving-systemsautonomous-drivingcameracollaborative-perceptioncomputer-visionconnected-and-automated-vehiclescooperative-driving-automationcooperative-perceptiondatasetlidar+10 more
Similar Tools
nuscenesargoverse
Keywords
multi-agent trustagent-to-agent evaluationcollaborative-perceptiondataset