Model Compression via Structural Pruning and Feature Distillation for Accurate Multi-Spectral Object Detection on Edge-Devices

Egor Poliakov, Van Tin Luu, Vu Hoang Tran, Ching Chun Huang

研究成果: Conference contribution同行評審

摘要

Multi-spectral infrared object detection across different infrared wavelengths is a challenging task. Although some full-sized object detection models, such as YOLOv4 and ScaledYOLO, may achieve good infrared object detection, they are resource-demanding and unsuitable for real-time detection on edge devices. Tiny versions for object detection are proposed to meet the practical requirement, but they usually sacrifice model accuracy and generalization for efficiency. We propose an accurate and efficient object detector capable of performing real-time inference under the hardware constraints of an edge device by leveraging structural pruning, feature distillation, and neural architecture search (NAS). The experiments on FLIR and multi-spectral object detection datasets show that our model achieves comparable mAP to full-sized models while having 14x times fewer parameters and 3.5x times fewer FLOPs. Our model can perform infrared detection well across different infrared wavelengths. The optimal CSPNet configurations of our detection network selected by NAS show that the resulting architectures outperform the baseline.

原文English
主出版物標題ICME 2022 - IEEE International Conference on Multimedia and Expo 2022, Proceedings
發行者IEEE Computer Society
ISBN(電子)9781665485630
DOIs
出版狀態Published - 2022
事件2022 IEEE International Conference on Multimedia and Expo, ICME 2022 - Taipei, Taiwan
持續時間: 18 7月 202222 7月 2022

出版系列

名字Proceedings - IEEE International Conference on Multimedia and Expo
2022-July
ISSN(列印)1945-7871
ISSN(電子)1945-788X

Conference

Conference2022 IEEE International Conference on Multimedia and Expo, ICME 2022
國家/地區Taiwan
城市Taipei
期間18/07/2222/07/22

指紋

深入研究「Model Compression via Structural Pruning and Feature Distillation for Accurate Multi-Spectral Object Detection on Edge-Devices」主題。共同形成了獨特的指紋。

引用此