https://scholars.lib.ntu.edu.tw/handle/123456789/607245
標題: | Online training refinement network and architecture design for stereo matching | 作者: | Wu Y.-S Wu S.-S Huang T LIANG-GEE CHEN |
關鍵字: | Device personalization;Layer fusion;Online training;Patch-based layer fusion;Refinement network;Stereo matching;Budget control;Online systems;Privacy by design;Architecture designs;Computation ability;Computer efficiency;Fusion techniques;Off-chip memory;State of the art;Stereo matching method;E-learning | 公開日期: | 2021 | 卷: | 2021-May | 來源出版物: | Proceedings - IEEE International Symposium on Circuits and Systems | 會議論文: | 53rd IEEE International Symposium on Circuits and Systems, ISCAS 2021 | 摘要: | Sending local data to cloud servers is vulnerable to user privacy, and its long update latency. Meanwhile, the state-of-the-art stereo matching method is still computation demanding, fine-tuning the whole model on-device is not a practicable solution because of the limited power budget and computation ability on edge devices. In this study, we propose a two-stage online stereo matching refinement system, using an additional light-weight network to learn the domain gap between local data and cloud training data. We define a load-gain ratio to evaluate computer efficiency. This refinement system has a much better load-gain ratio than fine-tune. (0.2 v.s. 35.7 operation overhead/accuracy gain) Nevertheless, we only disburse 0.2% of additional parameters and 0.7% additional computation as set by inference the stereo matching model. Thus, it would be a suitable choice for an online training scenario. With re-scheduling the training pipeline, we use a patch-based layer fusion technique and reduce the off-chip memory bandwidth by 97%. ? 2021 IEEE |
URI: | https://www.scopus.com/inward/record.uri?eid=2-s2.0-85109027731&doi=10.1109%2fISCAS51556.2021.9401468&partnerID=40&md5=3e2a45d0feb3299e07abc5c0c932669a https://scholars.lib.ntu.edu.tw/handle/123456789/607245 |
ISSN: | 02714310 | DOI: | 10.1109/ISCAS51556.2021.9401468 |
顯示於: | 電機工程學系 |
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。