DC Field | Value | Language |
---|---|---|
dc.contributor.author | Bruno, Andreis | ko |
dc.contributor.author | Park, Junhyeon | ko |
dc.contributor.author | Hwang, Sung Ju | ko |
dc.contributor.author | Kim, Minwoo | ko |
dc.date.accessioned | 2023-07-24T01:00:15Z | - |
dc.date.available | 2023-07-24T01:00:15Z | - |
dc.date.created | 2023-07-07 | - |
dc.date.created | 2023-07-07 | - |
dc.date.issued | 2018-07 | - |
dc.identifier.citation | 10th International Conference on Ubiquitous and Future Networks, ICUFN 2018, pp.64 - 69 | - |
dc.identifier.issn | 2165-8528 | - |
dc.identifier.uri | http://hdl.handle.net/10203/310764 | - |
dc.description.abstract | Advances in deep learning based object detection methods have achieve state-of-the-art detection accuracy in real-time using high-end GPUs. Their application to low-power computing systems (e.g. embedded GPUs on UAVs) is severely limited due to high computational requirements. We train a reinforcement learning agent to decide whether to perform object detection or tracking on a given image to maximize accuracy over execution time using visual differences between input frames. We validate our dynamic detection-tracking switching method on the Stanford Drone datasets for both detection accuracy and speed. Our model obtains comparable accuracy to the detector-only approach while obtaining 4x speedups. | - |
dc.language | English | - |
dc.publisher | IEEE Computer Society | - |
dc.title | Dynamic Detection-Tracking Switching | - |
dc.type | Conference | - |
dc.identifier.wosid | 000790260800019 | - |
dc.identifier.scopusid | 2-s2.0-85052541562 | - |
dc.type.rims | CONF | - |
dc.citation.beginningpage | 64 | - |
dc.citation.endingpage | 69 | - |
dc.citation.publicationname | 10th International Conference on Ubiquitous and Future Networks, ICUFN 2018 | - |
dc.identifier.conferencecountry | CS | - |
dc.identifier.conferencelocation | Prague | - |
dc.identifier.doi | 10.1109/ICUFN.2018.8436727 | - |
dc.contributor.localauthor | Hwang, Sung Ju | - |
dc.contributor.nonIdAuthor | Kim, Minwoo | - |
Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.