FRINK Linked Data Fragments server

Matches in SemOpenAlex for { <https://semopenalex.org/work/W3138117965> ?p ?o ?g. }

• W3138117965 abstract "Convolutional video models have an order of magnitude larger computational complexity than their counterpart image-level models. Constrained by computational resources, there is no model or training method that can train long video sequences end-to-end. Currently, the main-stream method is to split a raw video into clips, leading to incomplete fragmentary temporal information flow. Inspired by natural language processing techniques dealing with long sentences, we propose to treat videos as serial fragments satisfying Markov property, and train it as a whole by progressively propagating information through the temporal dimension in multiple steps. This progressive training (PGT) method is able to train long videos end-to-end with limited resources and ensures the effective transmission of information. As a general and robust training method, we empirically demonstrate that it yields significant performance improvements on different models and datasets. As an illustrative example, the proposed method improves SlowOnly network by 3.7 mAP on Charades and 1.9 top-1 accuracy on Kinetics with negligible parameter and computation overhead. Code is available at https://github.com/BoPang1996/PGT." @default.
• W3138117965 created "2021-03-29" @default.
• W3138117965 creator A5008809592 @default.
• W3138117965 creator A5010726528 @default.
• W3138117965 creator A5028272399 @default.
• W3138117965 creator A5058655480 @default.
• W3138117965 date "2021-03-21" @default.
• W3138117965 modified "2023-09-24" @default.
• W3138117965 title "PGT: A Progressive Method for Training Models on Long Videos" @default.
• W3138117965 cites W1485009520 @default.
• W3138117965 cites W1522734439 @default.
• W3138117965 cites W1686810756 @default.
• W3138117965 cites W1923404803 @default.
• W3138117965 cites W1947481528 @default.
• W3138117965 cites W2016053056 @default.
• W3138117965 cites W2057653135 @default.
• W3138117965 cites W2064675550 @default.
• W3138117965 cites W2095705004 @default.
• W3138117965 cites W2097117768 @default.
• W3138117965 cites W2150355110 @default.
• W3138117965 cites W2156303437 @default.
• W3138117965 cites W2161565164 @default.
• W3138117965 cites W2194775991 @default.
• W3138117965 cites W2337252826 @default.
• W3138117965 cites W2342662179 @default.
• W3138117965 cites W2518108298 @default.
• W3138117965 cites W2556967412 @default.
• W3138117965 cites W2618530766 @default.
• W3138117965 cites W2618799552 @default.
• W3138117965 cites W2619947201 @default.
• W3138117965 cites W2622263826 @default.
• W3138117965 cites W2706729717 @default.
• W3138117965 cites W2736115865 @default.
• W3138117965 cites W2746726611 @default.
• W3138117965 cites W2773514261 @default.
• W3138117965 cites W2799176631 @default.
• W3138117965 cites W2806331055 @default.
• W3138117965 cites W2809562466 @default.
• W3138117965 cites W2883275382 @default.
• W3138117965 cites W2887051120 @default.
• W3138117965 cites W2943833595 @default.
• W3138117965 cites W2955874753 @default.
• W3138117965 cites W2962841471 @default.
• W3138117965 cites W2962925365 @default.
• W3138117965 cites W2962934715 @default.
• W3138117965 cites W2963015194 @default.
• W3138117965 cites W2963091558 @default.
• W3138117965 cites W2963155035 @default.
• W3138117965 cites W2963230407 @default.
• W3138117965 cites W2963246338 @default.
• W3138117965 cites W2963315828 @default.
• W3138117965 cites W2963447094 @default.
• W3138117965 cites W2963524571 @default.
• W3138117965 cites W2963563276 @default.
• W3138117965 cites W2963645879 @default.
• W3138117965 cites W2963708869 @default.
• W3138117965 cites W2963722382 @default.
• W3138117965 cites W2964191259 @default.
• W3138117965 cites W2964241990 @default.
• W3138117965 cites W2964270168 @default.
• W3138117965 cites W2981578854 @default.
• W3138117965 cites W2988630963 @default.
• W3138117965 cites W2990152177 @default.
• W3138117965 cites W2990503944 @default.
• W3138117965 cites W2995684093 @default.
• W3138117965 cites W2998296459 @default.
• W3138117965 cites W3011765723 @default.
• W3138117965 cites W3018054553 @default.
• W3138117965 cites W3034429256 @default.
• W3138117965 cites W3034572008 @default.
• W3138117965 cites W3034895839 @default.
• W3138117965 cites W3035047011 @default.
• W3138117965 cites W3035285524 @default.
• W3138117965 cites W3035727180 @default.
• W3138117965 cites W3099322346 @default.
• W3138117965 cites W3109173645 @default.
• W3138117965 cites W3174207689 @default.
• W3138117965 doi "https://doi.org/10.48550/arxiv.2103.11313" @default.
• W3138117965 hasPublicationYear "2021" @default.
• W3138117965 type Work @default.
• W3138117965 sameAs 3138117965 @default.
• W3138117965 citedByCount "0" @default.
• W3138117965 crossrefType "posted-content" @default.
• W3138117965 hasAuthorship W3138117965A5008809592 @default.
• W3138117965 hasAuthorship W3138117965A5010726528 @default.
• W3138117965 hasAuthorship W3138117965A5028272399 @default.
• W3138117965 hasAuthorship W3138117965A5058655480 @default.
• W3138117965 hasBestOaLocation W31381179651 @default.
• W3138117965 hasConcept C111472728 @default.
• W3138117965 hasConcept C111919701 @default.
• W3138117965 hasConcept C11413529 @default.
• W3138117965 hasConcept C119857082 @default.
• W3138117965 hasConcept C121332964 @default.
• W3138117965 hasConcept C124101348 @default.
• W3138117965 hasConcept C138885662 @default.
• W3138117965 hasConcept C153294291 @default.
• W3138117965 hasConcept C154945302 @default.
• W3138117965 hasConcept C177264268 @default.
• W3138117965 hasConcept C179799912 @default.
• W3138117965 hasConcept C189950617 @default.

• next