Lightweight Improvements Based on RT-DETR

Tianjun Yi

doi:10.47297/taposatWSP2633-456921.20260706

PDF
导出
分享
收藏
专辑

Lightweight Improvements Based on RT-DETR

7卷第6期

浏览量: 0 Downloads: 0

Lightweight Improvements Based on RT-DETR
Lightweight Improvements Based on RT-DETR
School of Physics and Electronic Sciences, Changsha University of Science and Technology, Changsha, Hunan, 410000, China
2026年7卷第6期页码：102-108
网络首发：2026-08-18，

纸质出版：2026-06-30
DOI：10.47297/taposatWSP2633-456921.20260706
Accepted：

Scan QR Code

Tianjun Yi. (2026). Lightweight Improvements Based on RT-DETR. 科技理论与实践（英文版）, 7(6), 102-108.

Yi Tianjun.Lightweight Improvements Based on RT-DETR[J].Theory and Practice of Science and Technology,2026,07(06):102-108.
Tianjun Yi. (2026). Lightweight Improvements Based on RT-DETR. 科技理论与实践（英文版）, 7(6), 102-108. DOI： 10.47297/taposatWSP2633-456921.20260706.

Yi Tianjun.Lightweight Improvements Based on RT-DETR[J].Theory and Practice of Science and Technology,2026,07(06):102-108. DOI： 10.47297/taposatWSP2633-456921.20260706.

摘要

Abstract

This work suggests a lightweight approach that combines structural optimization with model compression to address the problems of the RT-DETR model

specifically its high computational cost

huge number of parameters

and difficulties in deploying it on edge devices. By incorporating the Ghost backbone network and CGC3 modules

we enhance inference efficiency and feature representation capabilities

and further reduce the model size through pruning

quantization

and model distillation. According to the experimental results

the modified model greatly reduces the number of model parameters while achieving a minor increase in mAP50 and AP50-95 on the VOC2012 dataset. Furthermore

when deployed on an edge computing platform

the model demonstrates a significant improvement in FPS

validating its effectiveness.

关键词

Keywords

references

Redmon, J., Divvala, S., Girshick, R., et al. 2016. You Only Look Once: Unified, Real-Time Object Detection. In: Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. Las Vegas. 779–788.

Carion, N., Massa, F., Synnaeve, G., et al. 2020. End-to-end object detection with transformers. In: European Conference on Computer Vision. Glasgow. 213–229.

Zhao, Y., Lv, W., Xu, S., Wei, J., Wang, G., Dang, Q., Liu, Y., Chen, J. 2024. DETRs Beat YOLOs on Real-time Object Detection. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. Seattle. 16965–16974.

Han, K., Wang, Y., Tian, Q., Guo, J., Xu, C., Xu, C. 2020. GhostNet: More Features From Cheap Operations. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. Seattle. 1580–1589.

Wu, T., Tang, S., Zhang, R., Zhang, Y. 2018. CGNet: A Light-weight Context Guided Network for Semantic Segmentation. In: Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. Salt Lake City. 11674–11683.

He, K., Zhang, X., Ren, S., Sun, J. 2016. Deep Residual Learning for Image Recognition. In: Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. Las Vegas. 770–778.

Liu, Z., Hao, Z., Han, K., Tang, Y., Wang, Y. 2024. GhostNetV3: Exploring the Training Strategies for Compact Models. arXiv preprint arXiv: 2404.11202 .

Li, H., Kadav, A., Durdanovic, I., Samet, H., Graf, H. P. 2017. Pruning Filters for Efficient ConvNets. In: International Conference on Learning Representations (ICLR). Toulon.

Han, S., Pool, J., Tran, J., Dally, W. 2015. Learning both Weights and Connections for Efficient Neural Networks. In: Advances in Neural Information Processing Systems 28. Montreal. 1135–1143.

Liu, Z., Li, J., Shen, Z., Huang, G., Yan, S., Zhang, C. 2017. Learning Efficient Convolutional Networks Through Network Slimming. In: Proceedings of the IEEE International Conference on Computer Vision. Venice. 2736–2744.

Jacob, B., Kligys, S., Chen, B., Zhu, M., Tang, M., Howard, A., Adam, H., Kalenichenko, D. 2018. Quantization and Training of Neural Networks for Efficient Integer-Arithmetic-Only Inference. In: Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. Salt Lake City. 2704–2713.

Hinton, G., Vinyals, O., Dean, J. 2015. Distilling the Knowledge in a Neural Network. arXiv preprint arXiv: 1503.02531 .

Romero, A., Ballas, N., Kahou, S. E., et al. 2015. FitNets: Hints for Thin Deep Nets. arXiv preprint arXiv: 1412.6550 .

Everingham, M., Van Gool, L., Williams, C. K. I., Winn, J., Zisserman, A. 2010. The PASCAL Visual Object Classes (VOC) Challenge. International Journal of Computer Vision. 88(2): 303–338.

文章被引用时，请邮件提醒。

Submit

Research on Road Pothole Detection System Based on Improved YOLOV11

Research on CAN Bus Injection Attack Detection Method for Internet of Vehicles Based on Deep Learning

A Review of Model Pruning Methods for Edge-Deployed MobileFaceNet

Physics Informed Neural Networks for PDE Coalescence

Research on Intelligent Early Warning and Prevention Control of Urban Building Fire Based on Multimodal Data Fusion

Related Author

Jialin Liu School of Information Engineering, North China University of Water Resources and Electric Power

Haoyu Liu College of computer science, Beijing Information Science&Technology University

Jianyong Yang School of Artificial Intelligence and Information Engineering, Yichun University

Zhang yu College of Science, China Jiliang University

Weihang Xue Shandong Technology and Business University

Zhaoyu Liu Shandong Technology and Business University

Yifan Li School of Physics, Sun Yat-sen University

Yihan Xu School of Health Science and Engineering, University of Shanghai for Science and Technology

Related Institution

School of Information Engineering, North China University of Water Resources and Electric Power

College of computer science, Beijing Information Science&Technology University

School of Artificial Intelligence and Information Engineering, Yichun University

College of Science, China Jiliang University

Shandong Technology and Business University

⁰