FogTTA: Online Test-Time Adaptation for Robust Transformer-based Object Detection in Foggy Weather

Key points

• Object detection models for autonomous driving commonly experience substantial performance drops when deployed under adverse weather due to the domain shift between training data and real-world operating conditions.1
• Using RF-DETR as the underlying object detector, FogTTA enables real-time adaptation to the streaming target domain without requiring source data or retraining.1
• The student model then learns from strongly augmented target sample using the Varifocal loss to mitigate pseudo-label noise.1
• Experiments show that FogTTA outperforms prior baselines, delivering improved detection accuracy and stability while maintaining real-time performance.1

Sources (1)

• [1]FogTTA: Online Test-Time Adaptation for Robust Transformer-based Object Detection in Foggy Weather
  arXiv (AI, ML, NLP, CV, robotics, multi-agent) · Oct 21, 11:57 AM

  To address this challenge, we introduce FogTTA, an online test-time adaptation framework designed to improve the robustness of Transformer-based object detectors in fog.

  Object detection models for autonomous driving commonly experience substantial performance drops when deployed under adverse weather due to the domain shift between training data and real-world operating conditions.

Extractive summary: sentences quoted from the sources.