Call For Paper - Upcoming Conferences
Research Article | Open Access | Download PDF
Volume 13 | Issue 7 | Year 2026 | Article Id. IJECE-V13I7P107 | DOI : https://doi.org/10.14445/23488549/IJECE-V13I7P107

SAFE-TTA: Safety-Aware Test-Time Adaptation for Real-Time Object Detection under Adverse Weather

Joseph Rish Simenthy, Pallavi Singh

Received Revised Accepted Published
21 May 2026 13 Jun 2026 30 Jun 2026 29 Jul 2026

Citation :

Joseph Rish Simenthy, Pallavi Singh, "SAFE-TTA: Safety-Aware Test-Time Adaptation for Real-Time Object Detection under Adverse Weather," International Journal of Electronics and Communication Engineering, vol. 13, no. 7, pp. 96-116, 2026. Crossref, https://doi.org/10.14445/23488549/IJECE-V13I7P107

Abstract

Weather conditions, including rain, snow, fog and low light, degrade the accuracy of modern object detectors by up to 40%. The same weather conditions are when reliable detection of vulnerable road users is needed most for enhanced road safety. Current TTA (Test Time Adaptation) approaches allow unconstrained gradient update, which may cause catastrophic forgetting, parameters diverge and invalidate certification. This paper presents SAFE-TTA, a framework that interprets Test-Time Adaptation as a Safety-Constrained Gradient-Free Optimization Problem. It includes four elements: (1) A SAFE adapter for low-rank residual feature transformations, with less than 9,000 trainable parameters in total; (2) Epistemic-Coupled Dynamic Envelopes (ECDE): Using Evidential Deep Learning, the adaptable area is dynamically reduced by evidentially derived limits on the increase of epistemic uncertainty. Under such extreme OOD scenarios, it will automatically revert back to the unadaptable baseline; (3) Constrained Evidential Particle Swarm Optimizer (CE-PSO); (4) Safety Monitor - providing analytically computable feature-space deviation bounds with automatic fail-safe reversion. under a zero-shot evaluation across four adverse-weather benchmarks (Foggy Cityscapes; BDD100K rain, snow, and night; ACDC), SAFE-TTA achieves 49.2 ± 0.4% average mAP@0.5 with a 31.4% relative reduction in pedestrian false negative rate at 42+ FPS with 1.8 ms per-frame overhead. All results are mean ± std over five independent runs. SAFE-TTA operates as an architecture-agnostic wrapper supporting both YOLOv11 and RT-DETR backbones.

Keywords

Test-Time Adaptation, Object detection, Autonomous navigation, Evidential Deep Learning, Epistemic uncertainty.

References

  1. Shaoqing Ren et al., “Faster R-CNN: Towards Real-Time Object Detection with Region Proposal Networks,” IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 39, no. 6, pp. 1137-1149, 2017.
    [    CrossRef] [Google Scholar] [Publisher Link]
  2. Christos Sakaridis, Dengxin Dai, and Luc Van Gool, “Semantic Foggy Scene Understanding with Synthetic Data,” International Journal of Computer Vision, vol. 126, pp. 973-992, 2018.
    [    CrossRef] [Google Scholar] [Publisher Link]
  3. Fisher Yu et al., “BDD100K: A Diverse Driving Dataset for Heterogeneous Multitask Learning,” 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), Seattle, WA, USA, pp. 2633-2642, 2020.
    [CrossRef] [Google Scholar] [Publisher Link]
  4. Christos Sakaridis, Dengxin Dai, and Luc Van Gool, “ACDC: The Adverse Conditions Dataset with Correspondences for Semantic Driving Scene Understanding,” 2021 IEEE/CVF International Conference on Computer Vision (ICCV), Montreal, QC, Canada, pp. 10745-10755, 2021.
    [    CrossRef] [Google Scholar] [Publisher Link]
  5. Dequan Wang et al., “Tent: Fully Test-Time Adaptation by Entropy Minimization,” arXiv preprint, pp. 1-15, 2020.
    [    CrossRef] [Google Scholar] [Publisher Link]
  6. Qin Wang et al., “Continual Test-Time Domain Adaptation,” 2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), New Orleans, LA, USA, pp. 7191-7201, 2022.
    [CrossRef] [Google Scholar] [Publisher Link]
  7. Shuaicheng Niu et al., “Towards Stable Test-Time Adaptation in Dynamic Wild World,” arXiv preprint, pp. 1-27, 2023.
    [    CrossRef] [Google Scholar] [Publisher Link]
  8. Marvin Zhang, Sergey Levine, and Chelsea Finn, “MEMO: Test Time Robustness via Adaptation and Augmentation,” Advances in Neural Information Processing Systems, vol. 35, pp. 38629-38642, 2022.
    [    CrossRef] [Google Scholar] [Publisher Link]
  9. Shuaicheng Niu et al., “Efficient Test-Time Model Adaptation without Forgetting,” Proceedings of the 39th International Conference on Machine Learning (ICML), vol. 162, pp. 16888-16905, 2022.
    [CrossRef] [Google Scholar] [Publisher Link]
  10. Longhui Yuan, Binhui Xie, and Shuang Li, “Robust Test-Time Adaptation in Dynamic Scenarios,” 2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), Vancouver, BC, Canada, pp. 15922-15932, 2023.
    [    CrossRef] [Google Scholar] [Publisher Link]
  11. Murat Sensoy, Lance Kaplan, and Melih Kandemir, “Evidential Deep Learning to Quantify Classification Uncertainty,” Advances in Neural Information Processing Systems, vol. 31, 2018.
    [    CrossRef] [Google Scholar] [Publisher Link]
  12. Junha Song et al., “EcoTTA: Memory-Efficient Continual Test-Time Adaptation via Self-Distilled Regularization,” 2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), Vancouver, BC, Canada, pp. 11920-11929, 2023.
    [    CrossRef] [Google Scholar] [Publisher Link]
  13. Ziyang Chen et al., “Each Test Image Deserves A Specific Prompt: Continual Test-Time Adaptation for 2D Medical Image Segmentation,” 2024 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), Seattle, WA, USA, pp. 11184-11193, 2024.
    [    CrossRef] [Google Scholar] [Publisher Link]
  14. Jiaming Liu et al., “Continual-MAE: Adaptive Distribution Masked Autoencoders for Continual Test-Time Adaptation,” 2024 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), Seattle, WA, USA, pp. 28653-28663, 2024.
    [    CrossRef] [Google Scholar] [Publisher Link]
  15. Xinru Meng et al., “SPARNet: Continual Test-Time Adaptation via Sample Partitioning Strategy and Anti-Forgetting Regularization,” arXiv preprint, pp. 1-8, 2025.
    [    CrossRef] [Google Scholar] [Publisher Link]
  16. Shilei Cao et al., “Exploring Test-Time Adaptation for Object Detection in Continually Changing Environments,” arXiv preprint, vol. 14, no. 8, pp. 1-18, 2024.
    [    Google Scholar] [Publisher Link]
  17. Atif Belal et al., “VLOD-TTA: Test-Time Adaptation of Vision-Language Object Detectors,” arXiv preprint, pp. 1-34, 2025.
    [    CrossRef] [Google Scholar] [Publisher Link]
  18. Lihua Zhou et al., “Bayesian Test-Time Adaptation for Object Recognition and Detection with Vision-Language Models,” arXiv preprint, pp. 1-18, 2025.
    [    CrossRef] [Google Scholar] [Publisher Link]
  19. Boyi Li et al., “AOD-Net: All-in-One Dehazing Network,” 2017 IEEE International Conference on Computer Vision (ICCV), Venice, Italy, pp. 4770-4778, 2017.
    [    CrossRef] [Google Scholar] [Publisher Link]
  20. Xueyang Fu et al., “Removing Rain from Single Images via a Deep Detail Network,” 2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), Honolulu, HI, USA, pp. 1715-1723, 2017.
    [    CrossRef] [Google Scholar] [Publisher Link]
  21. Jeya Maria Jose Valanarasu, Rajeev Yasarla, and Vishal M. Patel, “TransWeather: Transformer-based Restoration of Images Degraded by Adverse Weather Conditions,” 2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), New Orleans, LA, USA, pp. 2343-2353, 2022.
    [    CrossRef] [Google Scholar] [Publisher Link]
  22. Yuhua Chen et al., “Domain Adaptive Faster R-CNN for Object Detection in the Wild,” 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition, Salt Lake City, UT, USA, pp. 3339-3348, 2018.
    [    CrossRef] [Google Scholar] [Publisher Link]
  23. Wuyang Li, Xinyu Liu, and Yixuan Yuan, “SIGMA: Semantic-Complete Graph Matching for Domain Adaptive Object Detection,” 2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), New Orleans, LA, USA, pp. 5281-5290, 2022.
    [    CrossRef] [Google Scholar] [Publisher Link]
  24. Kunyi Wang, and Yaohua Zhao, “Improving Object Detection in Challenging Weather for Autonomous Driving via Adversarial Image Translation,” Plos One, vol. 20, no. 10, pp. 1-18, 2025.
    [    CrossRef] [Google Scholar] [Publisher Link]
  25. Liye Jiang et al., “YOLO-DH: Robust Object Detection for Autonomous Vehicles in Adverse Weather,” Electronics, vol. 14, no. 22, pp. 1-18, 2025.
    [    CrossRef] [Google Scholar] [Publisher Link]
  26. Qi Fan et al., “Robust Object Detection with Domain-Invariant Training and Continual Test-Time Adaptation,” International Journal of Computer Vision, vol. 133, no. 10, pp. 6768-6793, 2025.
    [    CrossRef] [Google Scholar] [Publisher Link]
  27. Nasser Aloufi, Abdulaziz Alnori, and Abdullah Basuhail, “Enhancing Autonomous Vehicle Perception in Adverse Weather: A Multi Objectives Model for Integrated Weather Classification and Object Detection,” Electronics, vol. 13, no. 15, pp. 1-20, 2024.
    [    CrossRef] [Google Scholar] [Publisher Link]
  28. Debasis Kumar, and Naveed Muhammad, “Object Detection in Adverse Weather for Autonomous Driving through Data Merging and YOLOv8,” Sensors, vol. 23, no. 20, pp. 1-29, 2023.
    [    CrossRef] [Google Scholar] [Publisher Link]
  29. İbrahim Özcan, Yusuf Altun, and Cevahir Parlak, “Improving YOLO Detection Performance of Autonomous Vehicles in Adverse Weather Using Metaheuristic Algorithms,” Applied Sciences, vol. 14, no. 13, pp. 1-25, 2024.
    [    CrossRef] [Google Scholar] [Publisher Link]
  30. Alok Kumar Tiwari, Manisha Pattanaik, and G. K. Sharma, “Low-Light DEtection TRansformer (LDETR): Object Detection in Low-light and Adverse Weather Conditions,” Multimedia Tools and Applications, vol. 83, no. 36, pp. 84231-84248, 2024.
    [    CrossRef] [Google Scholar] [Publisher Link]
  31. Wenyu Liu et al., “Image-Adaptive YOLO for Object Detection in Adverse Weather Conditions,” Proceedings of the AAAI Conference on Artificial Intelligence, vol. 36, no. 2, pp. 1792-1800, 2022.
    [    CrossRef] [Google Scholar] [Publisher Link]
  32. Balaji Lakshminarayanan, Alexander Pritzel, and Charles Blundell, “Simple and Scalable Predictive Uncertainty Estimation using Deep Ensembles,” Advances in Neural Information Processing Systems, vol. 30, pp. 1-15, 2017.
    [    CrossRef] [Google Scholar] [Publisher Link]
  33. Yarin Gal, and Zoubin Ghahramani, “Dropout as a Bayesian Approximation: Representing Model Uncertainty in Deep Learning,” Proceedings of the 33rd International Conference on Machine Learning, vol. 48, pp. 1050-1059, 2016.
    [    CrossRef] [Google Scholar] [Publisher Link]
  34. David Hall et al., “Probabilistic Object Detection: Definition and Evaluation,” 2020 IEEE Winter Conference on Applications of Computer Vision (WACV), Snowmass, CO, USA, pp. 1020-1029, 2020.
    [    CrossRef] [Google Scholar] [Publisher Link]
  35. Danut-Vasile Giurgi et al., “Intelligent Vehicles Semantic Segmentation Using Evidential Deep Learning,” Upper Rhine Artificial Intelligence Symposium, pp. 40-49, 2023.
    [    CrossRef] [Google Scholar] [Publisher Link]
  36. Wentao Bao, Qi Yu, and Yu Kong, “Evidential Deep Learning for Open Set Action Recognition,” 2021 IEEE/CVF International Conference on Computer Vision (ICCV), Montreal, QC, Canada, pp. 13329-13338, 2021.
    [    CrossRef] [Google Scholar] [Publisher Link]
  37. Young-Jin Park, Carson Sobolewski, and Navid Azizan, “Uncertainty Quantification in Detection Transformers: Object-Level Calibration and Image-Level Reliability,” IEEE Transactions on Pattern Analysis and Machine Intelligence, pp. 1-14, 2026.
    [CrossRef] [Google Scholar] [Publisher Link]
  38. Monish R. Nallapareddy et al., “EvCenterNet: Uncertainty Estimation for Object Detection using Evidential Learning,” 2023 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), Detroit, MI, USA, pp. 5699-5706, 2023.
    [    CrossRef] [Google Scholar] [Publisher Link]
  39. Taeseong Yoon, and Heeyoung Kim, “Uncertainty Estimation by Flexible Evidential Deep Learning,” arXiv preprint, pp. 1-41, 2026.
    [CrossRef] [Google Scholar] [Publisher Link]
  40. Taeseong Yoon, and Heeyoung Kim, “Uncertainty Estimation by Density Aware Evidential Deep Learning,” arXiv preprint, pp. 1-27, 2024.
    [    CrossRef] [Google Scholar] [Publisher Link]
  41. J. Kennedy, and R. Eberhart, “Particle Swarm Optimization,” Proceedings of ICNN'95 - International Conference on Neural Networks, Perth, WA, Australia, vol. 4, pp. 1942-1948, 1995.
    [    CrossRef] [Google Scholar] [Publisher Link]
  42. Guy Katz et al., “Reluplex: An Efficient SMT Solver for Verifying Deep Neural Networks,” International Conference on Computer-Aided Verification (CAV), Lecture Notes in Computer Science, vol. 10426, pp. 97-117, 2017.
    [    CrossRef] [Google Scholar] [Publisher Link]
  43. Weiming Xiang, “RunTime Safety Monitoring of Neural-Network-Enabled Dynamical Systems,” IEEE Transactions on Cybernetics, vol. 52, no. 9, pp. 9587-9596, 2022.
    [    CrossRef] [Google Scholar] [Publisher Link]
  44. Dengxin Dai et al., “Curriculum Model Adaptation with Synthetic and Real Data for Semantic Foggy Scene Understanding,” arXiv preprint, pp. 1-22, 2019.
    [CrossRef] [Google Scholar] [Publisher Link]
  45. Kai Yang et al., “Uncertainties in Onboard Algorithms for Autonomous Vehicles: Challenges, Mitigation, and Perspectives,” IEEE Transactions on Intelligent Transportation Systems, vol. 24, no. 9, pp. 8963-8987, 2023.
    [    CrossRef] [Google Scholar] [Publisher Link]
  46. Zhaohui Liu et al., “Robust Object Detection in Adverse Weather Conditions: ECL-YOLOv11 for Automotive Vision Systems,” Sensors, vol. 26, no. 1, pp. 1-24, 2026.
    [    CrossRef] [Google Scholar] [Publisher Link]
  47. Youngjun Song, Hyeongyu Kim, and Dosik Hwang, “CD-Buffer: Complementary Dual-Buffer Framework for Test-Time Adaptation in Adverse Weather Object Detection,” Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pp. 15050-15059, 2026.
    [    CrossRef] [Google Scholar] [Publisher Link]
  48. Nkalubo Lenard Byenkya, Nakibuule Rose, and Okila Nixson, “Adaptive Object Detection in Dynamic Road Environments Using Optimized YOLOv11 with Continuous Learning,” Discover Artificial Intelligence, vol. 6, pp. 1-31, 2026.
    [    CrossRef] [Google Scholar] [Publisher Link]
  49. Yian Zhao et al., “DETRs Beat YOLOs on Real-time Object Detection,” 2024 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), Seattle, WA, USA, pp. 16965-16974, 2024.
    [    CrossRef] [Google Scholar] [Publisher Link]