https://scholars.lib.ntu.edu.tw/handle/123456789/607502
標題: | Exploring the power of lightweight YOLOv4 | 作者: | Wang C.-Y Liao H.-Y.M Yeh I.-H Lin Y.-L. YUNG-YU CHUANG |
關鍵字: | Deep learning;Electric power utilization;Low power electronics;Object detection;System-on-chip;Analysis of data;Data space;Learning methods;Low-costs;Low-power consumption;Lower-power consumption;Neural-networks;Power;Projection space;Space models;Network architecture | 公開日期: | 2021 | 卷: | 2021-October | 起(迄)頁: | 779-788 | 來源出版物: | Proceedings of the IEEE International Conference on Computer Vision | 摘要: | Research on deep learning has always had two main streams: (1) design a powerful network architecture and train it with existing learning methods to achieve the best results, and (2) design better learning methods so that the existing network architecture can achieve the best capbility after training. In recent years, because mobile device has become popular, the requirement of low power consumption becomes a must. Under the requirement of low power consumption, we hope to design low-cost lightweight networks that can be effectively deployed at the edge, while it must have enough resources to be used and the inference speed must be fast enough. In this work, we set a very ambitious goal of exploring the power of lightweight neural networks. We utilize the analysis of data space, model's representational capacity, and knowledge projection space to construct an automated machine learning pipeline. Through this mechanism, we systematically derive the most suitable knowledge projection space between the data and the model. Our method can indeed automatically find learning strategies suitable for the target model and target application through exploration. Experiment results show that the proposed method can significantly enhance the accuracy of lightweight neural networks for object detection. We directly apply the lightweight model trained by our proposed method to a Jetson Xavier NX embedded module and a Kneron KL720 edge AI SoC as system solutions. ? 2021 IEEE. Research on deep learning has always had two main streams: (1) design a powerful network architecture and train it with existing learning methods to achieve the best results, and (2) design better learning methods so that the existing network architecture can achieve the best capbility after training. In recent years, because mobile device has become popular, the requirement of low power consumption becomes a must. Under the requirement of low power consumption, we hope to design low-cost lightweight networks that can be effectively deployed at the edge, while it must have enough resources to be used and the inference speed must be fast enough. In this work, we set a very ambitious goal of exploring the power of lightweight neural networks. We utilize the analysis of data space, model's representational capacity, and knowledge projection space to construct an automated machine learning pipeline. Through this mechanism, we systematically derive the most suitable knowledge projection space between the data and the model. Our method can indeed automatically find learning strategies suitable for the target model and target application through exploration. Experiment results show that the proposed method can significantly enhance the accuracy of lightweight neural networks for object detection. We directly apply the lightweight model trained by our proposed method to a Jetson Xavier NX embedded module and a Kneron KL720 edge AI SoC as system solutions. © 2021 IEEE. |
URI: | https://www.scopus.com/inward/record.uri?eid=2-s2.0-85123047408&doi=10.1109%2fICCVW54120.2021.00092&partnerID=40&md5=d4af54ad2cc79338cfed47f77b316383 https://scholars.lib.ntu.edu.tw/handle/123456789/607502 |
ISSN: | 15505499 | DOI: | 10.1109/ICCVW54120.2021.00092 | SDG/關鍵字: | Deep learning; Electric power utilization; Low power electronics; Object detection; System-on-chip; Analysis of data; Data space; Learning methods; Low-costs; Low-power consumption; Lower-power consumption; Neural-networks; Power; Projection space; Space models; Network architecture |
顯示於: | 資訊工程學系 |
在 IR 系統中的文件,除了特別指名其著作權條款之外,均受到著作權保護,並且保留所有的權利。