Journal of Technology and Informatics
Condition-Aware Cross-Scale Transformers for Object Detection under Compound Adverse Environments
DOI:
https://doi.org/10.37802/Keywords:
adverse weather detection, autonomous driving, condition-aware adaptation, curriculum learning, object detection, transformer architectureAbstract
Object detection under adverse environmental conditions remains a critical challenge for autonomous driving systems, with performance degrading by over 40% under dense fog, heavy precipitation, or nighttime scenarios. Existing approaches either specialize for single degradation types or remain fundamentally condition-agnostic, failing to address the compound nature of real-world adverse conditions. We present a Condition-Aware Cross-Scale Transformer (CACST), a unified framework integrating explicit environmental condition estimation, adaptive image preprocessing, and cross-scale feature interaction. Our approach introduces: (1) a condition estimation module quantifying degradation parameters via supervised and contrastive learning; (2) a gated image adaptation mechanism with learnable ISP operations dynamically modulated by estimated conditions; and (3) an Enhanced Density-aware Cross-Scale Transformer (EDST) performing global attention across feature pyramid scales with spatial reliability weighting. Comprehensive experiments across five benchmarks demonstrate state-of-the-art performance: CACST achieves +4.1% mAP on ACDC (47.8% vs. 43.7%), +3.2% on Foggy Cityscapes (49.7% vs. 46.5%), and +6.7% on ExDark (74.9% vs. 68.2%) compared to the strongest specialized baseline, maintaining balanced accuracy across all conditions
Downloads
References
[1] K. He, X. Zhang, S. Ren, and J. Sun, "Deep residual learning for image recognition," in Proc. IEEE Conf. Comput. Vis. Pattern Recognit. (CVPR), Las Vegas, NV, USA, 2016, pp. 770–778.
[2] S. Ren, K. He, R. Girshick, and J. Sun, "Faster R-CNN: Towards real-time object detection with region proposal networks," in Adv. Neural Inf. Process. Syst. (NeurIPS), vol. 28, 2015.
[3] T.-Y. Lin, P. Goyal, R. Girshick, K. He, and P. Dollár, "Focal loss for dense object detection," in Proc. IEEE Int. Conf. Comput. Vis. (ICCV), Venice, Italy, 2017, pp. 2980–2988.
[4] H. Gupta, O. Kotlyar, H. Andreasson, and A. J. Lilienthal, "Robust object detection in challenging weather conditions," in Proc. IEEE/CVF Winter Conf. Appl. Comput. Vis. (WACV), 2024, pp. 7471–7482.
[5] Y. Chae, H. Kim, and K.-J. Yoon, "Towards robust 3D object detection with LiDAR and 4D radar fusion in various weather conditions," in Proc. IEEE/CVF Conf. Comput. Vis. Pattern Recognit. (CVPR), 2024, pp. 15162–15172.
[6] E. O. Appiah and S. Mensah, "Object detection in adverse weather condition for autonomous vehicles," Multimedia Tools Appl., vol. 83, pp. 28235–28261, 2024, doi: 10.1007/s11042-023-16836-y.
[7] Y. Zhao et al., "DETRs beat YOLOs on real-time object detection," in Proc. IEEE/CVF Conf. Comput. Vis. Pattern Recognit. (CVPR), 2024, pp. 16788–16797.
[8] W. Lv et al., "RT-DETRv2: Improved baseline with bag-of-freebies for real-time detection transformer," arXiv preprint arXiv:2407.17140, 2024.
[9] Y. Wang, Y. Lu, and Y. Qiu, "Gated image-adaptive network for driving-scene object detection under nighttime conditions," Multimedia Syst., vol. 31, no. 23, pp. 1–18, 2025.
[10] F. Oshasha et al., "EDCST-Rain: Enhanced Density-Aware Cross-Scale Transformer for Robust Object Classification Under Diverse Rainfall Conditions," J. Appl. Inform. Comput. (JAIC), vol. 10, no. 1, pp. 33–45, Feb. 2026, doi: 10.30871/jaic.v10i1.11590.
[11] N. C. Imanuela Natun, M. A. Santhia, and Y. R. Kaesmetan, "Identifikasi Pengenalan Wajah Berdasarkan Jenis Kelamin Menggunakan Metode Convolutional Neural Network (CNN)," J. Technol. Inform. (JoTI), vol. 6, no. 1, pp. 50–57, Oct. 2024, doi: 10.37802/joti.v6i1.694.
[12] F. A. Subkhi, M. Fuad, S. Wahyuni, A. I. Sudianto, V. T. Widyaningrum, and A. Dafid, "A Real-Time Human-Drone Interaction System for Cornfield Perimeter Monitoring Using Hand Gesture Control," J. Technol. Inform. (JoTI), vol. 8, no. 1, pp. 73–83, Apr. 2026, doi: 10.37802/joti.v8i1.1314.
[13] W. I. Kusumawati, H. Pratikno, and Y. P. Admaja, "Sistem Penghitung Jumlah Pengunjung Restoran Menggunakan Kamera Berbasis Single Shot Detector (SSD)," J. Technol. Inform. (JoTI), vol. 3, no. 1, pp. 19–26, Oct. 2021, doi: 10.37802/joti.v3i1.197.
[14] A. Murdiono, M. Fuad, H. Budiarto, F. Umam, V. T. Widyaningrum, and A. I. Sudianto, "Brush-shaped Motion Gesture of UGV Using Hand Gesture Recognition," J. Technol. Inform. (JoTI), vol. 8, no. 1, pp. 25–36, Apr. 2026, doi: 10.37802/joti.v8i1.1315.
[15] N. Carion et al., "End-to-end object detection with transformers," in Proc. Eur. Conf. Comput. Vis. (ECCV), 2020, pp. 213–229.
[16] F. Wang, X. Yang, and J. Wei, "YOLOv8-STE: Enhancing object detection performance under adverse weather conditions," Electronics, vol. 13, no. 24, p. 5049, 2024, doi: 10.3390/electronics13245049.
[17] J. Li, H. Wang, X. Zhang, and W. Liu, "MCD-Net: Multi-component decomposition network for object detection under adverse weather conditions," Pattern Recognit., vol. 160, p. 111180, 2026.
[18] L. Jiao et al., "CW-DETR: Cross-weather detection transformer with adaptive dense bi-level feature pyramid network," in Proc. IEEE/CVF Winter Conf. Appl. Comput. Vis. (WACV), 2026.
[19] B. Zhang et al., "MDAT-YOLO: Multi-dimensional attention transformer for object detection under adverse weather conditions," Sci. Rep., vol. 16, p. 7831, 2026.
[20] T.-Y. Lin et al., "Feature pyramid networks for object detection," in Proc. IEEE Conf. Comput. Vis. Pattern Recognit. (CVPR), 2017, pp. 2117–2125.
[21] M. Cordts et al., "The cityscapes dataset for semantic urban scene understanding," in Proc. IEEE Conf. Comput. Vis. Pattern Recognit. (CVPR), 2016, pp. 3213–3223.
[22] C. Sakaridis, D. Dai, and L. Van Gool, "Semantic foggy scene understanding with synthetic data," Int. J. Comput. Vis., vol. 126, no. 9, pp. 973–992, 2018, doi: 10.1007/s11263-018-1072-8.
[23] C. Sakaridis, D. Dai, and L. Van Gool, "ACDC: The adverse conditions dataset with correspondences," in Proc. IEEE/CVF Int. Conf. Comput. Vis. (ICCV), 2021, pp. 10765–10775.
[24] Y. P. Loh and C. S. Chan, "Getting to know low-light images with the exclusively dark dataset," Comput. Vis. Image Underst., vol. 178, pp. 30–42, 2019, doi: 10.1016/j.cviu.2018.10.010.
[25] M. A. Kenk and M. Hassaballah, "DAWN: Vehicle detection in adverse weather nature dataset," arXiv preprint arXiv:2008.05402, 2020.
[26] Z. Tian, C. Shen, H. Chen, and T. He, "FCOS: Fully convolutional one-stage object detection," in Proc. IEEE/CVF Int. Conf. Comput. Vis. (ICCV), 2019, pp. 9627–9636.
[27] W. Liu et al., "IA-YOLO: Image-adaptive YOLO for small-object detection in low-illumination and adverse weather," arXiv preprint arXiv:2212.01528, 2022.
[28] Z. Xu et al., "D-YOLO: A robust framework for object detection in adverse conditions via decoupled feature alignment," Comput. Vis. Image Underst., vol. 248, p. 104118, 2024, doi: 10.1016/j.cviu.2024.104118.
[29] Z. Cui et al., "Multitask AET with orthogonal tangent regularity for dark object detection," in Proc. IEEE/CVF Int. Conf. Comput. Vis. (ICCV), 2021, pp. 2553–2562.
[30] Z. Li et al., "SAMFusion: Sensor-adaptive multimodal fusion for 3D object detection in adverse weather," in Proc. Eur. Conf. Comput. Vis. (ECCV), Springer, 2024, pp. 464–481.
[31] F. S. Kyara, A. Aryanti, and R. A. Halimatussa'diyah, "Sistem Pengenalan Wajah Real-Time Menggunakan YOLOv7 untuk Akses Gedung TVRI Palembang Berbasis Web," J. Technol. Inform. (JoTI), vol. 7, no. 2, pp. 147–160, Oct. 2025, doi: 10.37802/joti.v7i2.1063.
[32] A. T. P. Tata, I. Salamah, and M. M. Rose, "Implementasi Sistem Keamanan Brankas Berbasis Face Recognition Menggunakan Algoritma YOLO dengan Verifikasi Fingerprint," J. Technol. Inform. (JoTI), vol. 7, no. 2, pp. 161–174, Oct. 2025, doi: 10.37802/joti.v7i2.1072.
[33] M. R. D. Ulhaq, M. A. Zaidan, and D. Firdaus, "Pengenalan Ekspresi Wajah Secara Real-Time Menggunakan Metode SSD Mobilenet Berbasis Android," J. Technol. Inform. (JoTI), vol. 5, no. 1, pp. 48–52, Oct. 2023, doi: 10.37802/joti.v5i1.387.
