DC Field | Value | Language |
---|---|---|
dc.contributor.author | Han, DongJun | ko |
dc.contributor.author | Park, JungWuk | ko |
dc.contributor.author | Ham, Seokil | ko |
dc.contributor.author | Lee, Namjin | ko |
dc.contributor.author | Moon, Jaekyun | ko |
dc.date.accessioned | 2022-11-28T08:04:38Z | - |
dc.date.available | 2022-11-28T08:04:38Z | - |
dc.date.created | 2022-11-28 | - |
dc.date.issued | 2022-06-20 | - |
dc.identifier.citation | 2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops, CVPRW 2022 | - |
dc.identifier.uri | http://hdl.handle.net/10203/301180 | - |
dc.description.abstract | Multi-exit architecture is a promising solution that can make adaptive predictions via early exits, depending on the current test-time budget which may vary over time in practice (e.g., self-driving cars with dynamically changing speeds). Compared to the previous works where each block is optimized to minimize the losses of all exits simultaneously, we propose a new method for training multi-exit architectures by imposing different objectives to individual blocks. Our key idea is to design block-dependent losses based on grouping and overlapping strategies, which enables each k-th block to focus more on reducing the loss of the adjacent k-th exit while not degrading the prediction performance at later exits. This improves the prediction performance at the earlier exits, making our scheme to be more suitable for low-latency applications with a tight test-time budget. Experimental results on both image classification and semantic segmentation confirm the advantage of our approach for anytime prediction. | - |
dc.language | English | - |
dc.publisher | IEEE Computer Society | - |
dc.title | Training Multi-Exit Architectures via Block-Dependent Losses for Anytime Inference | - |
dc.type | Conference | - |
dc.type.rims | CONF | - |
dc.citation.publicationname | 2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops, CVPRW 2022 | - |
dc.identifier.conferencecountry | US | - |
dc.identifier.conferencelocation | New Orleans, LA | - |
dc.contributor.localauthor | Moon, Jaekyun | - |
dc.contributor.nonIdAuthor | Ham, Seokil | - |
dc.contributor.nonIdAuthor | Lee, Namjin | - |
Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.