Enhancing Object Detection in Autonomous Driving Under Extreme Conditions: A Comprehensive Study of Deep Learning Techniques
DOI:
https://doi.org/10.61173/xqj1a096Keywords:
Autonomous driving, object detection, generative adversarial networks, adverse weatherAbstract
This testimonial deals with the issue of efficiency destruction in deep learning-driven object detection models for autonomous driving systems under severe and intricate conditions. This problem is important for boosting the safety and security and dependability of autonomous driving, especially in unfavorable weather conditions and low-light conditions. The paper methodically examines numerous techniques to boost the effectiveness of object detection models in difficult atmospheres. Trick methods consist of making use of Generative Adversarial Networks (GANs) to produce artificial wet information, improving the training procedure by imitating varied wet circumstances that the model might come across in real-world conditions. By doing so, the model ends up being extra experienced at managing the aesthetic distortions triggered by hefty rainfall. Moreover, the blend of multisensor information, such as incorporating electronic camera photos with radar and LIDAR information, makes up for the restrictions of specific sensing units by supplying added spatial and range info, which is much less influenced by unfavorable climate. The introduction of spatial attention mechanisms in network styles likewise plays a substantial duty, enabling models to concentrate on one of the most pertinent locations of a picture, hence maximizing detection efficiency in complicated roadway situations. In addition, for nighttime driving, the paper discovers the application of sophisticated picture correction strategies. Using high-sensitivity cams and deep learning-based techniques for decreasing headlight glare even more adds to enhanced object detection under tough illumination conditions. The findings show that these approaches efficiently boost the toughness of object detection models in intricate settings.References
[1] Redmon J. YOLOv3: An incremental improvement. arXiv preprint arXiv:1804.02767. 2018.
[2] Ren S, He K, Girshick R, Sun J. Faster R-CNN: Towards real-time object detection with region proposal networks. IEEE Trans Pattern Anal Mach Intell. 2016;39(6):1137-49.
[3] Liu W, Anguelov D, Erhan D, Szegedy C, Reed S, Fu CY, Berg AC. SSD: Single shot multibox detector. In: Computer Vision–ECCV 2016: 14th European Conference, Amsterdam, The Netherlands, October 11–14, 2016, Proceedings, Part I. Springer International Publishing; 2016. p. 21-37.
[4] Zhang X, Wang W, Wu Z, Liang Y. Multi-source fusion for object detection in adverse weather conditions. IEEE Trans Intell Transp Syst. 2020;22(2):884-95.
[5] Li H, Fan X, Wang Y. Augmentation techniques for deep learning in autonomous driving under adverse weather conditions. In: Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition Workshops. 2020. p. 141-50.
[6] Kumar R, Gupta S. A robust fusion of CNN and LIDAR for enhanced autonomous driving in complex environments. J Field Robot. 2021;38(6):928-42.
[7] Zhang C, Gao J, Tang Y. Rainy condition adaptation for object detection in autonomous driving. Pattern Recogn Lett. 2021;140:1-7.
[8] Li W, Xie X, Fu K. Deep convolutional networks for image defogging in autonomous vehicles. IEEE Trans Vehicular Technol. 2019;68(7):6650-61.
[9] Tan M, Le QV. EfficientNet: Rethinking model scaling for convolutional neural networks. In: Proceedings of the 36th International Conference on Machine Learning. 2021. p. 6105- 14.
[10] Chen S, Liu X, Tang R. Multi-sensor data fusion for improved object detection in foggy conditions. Sensors. 2022;22(5):1867.
[11] Zhou Z, Xu Y, Wu Q. Low-light image enhancement for autonomous driving: A survey. IEEE Trans Intell Transp Syst. 2020;21(11):4565-80.
[12] Yang L, Wu H, Yang G. Wide dynamic range image processing in low-light conditions for autonomous driving. IEEE Access. 2022;10:43552-62.
[13] Tang H, Liu Y. Glare reduction techniques for nighttime driving: A deep learning approach. J Opt Soc Am A. 2021;38(9):1256-65.
[14] Chen X, Ma H, Wang J. Spatial attention mechanisms for improving object detection in complex road scenarios. IEEE Trans Intell Transp Syst. 2020;22(3):1451-62.
[15] Wang Y, Li S, Lu H. Spatial attention for urban driving scenarios: Improving detection in complex environments. Pattern Recognit. 2020;107:107469.
[16] Peng Z, Zhu X. Data augmentation strategies for autonomous driving in complex road scenarios. IEEE Trans Image Process. 2021;30:4625-36.
[17] Doshi-Velez F, Kim B. Towards a rigorous science of interpretable machine learning. arXiv preprint arXiv:1702.08608. 2017.
[18] Torralba A, Efros AA. Unbiased look at dataset bias. In: CVPR 2011. 2011. p. 1521-8.
[19] Han S, Mao H, Dally WJ. Deep compression: Compressing deep neural networks with pruning, trained quantization, and Huffman coding. arXiv preprint arXiv:1510.00149. 2015.
[20] Ribeiro MT, Singh S, Guestrin C. “Why should I trust you?” Explaining the predictions of any classifier. In: Proceedings of the 22nd ACM SIGKDD international conference on knowledge detection and data mining. 2016. p. 1135-44.
[21] Pan SJ, Yang Q. A survey on transfer learning. IEEE Trans Knowl Data Eng. 2010;22(10):1345-59.
[22] Satyanarayanan M. The emergence of edge computing. Computer. 2017;50(1):30-9.
Downloads
Published
Issue
Section
License
Copyright (c) 2024 by the authors.
This work is licensed under a Creative Commons Attribution 4.0 International License.
