Matches in SemOpenAlex for { <https://semopenalex.org/work/W4385800567> ?p ?o ?g. }
Showing items 1 to 83 of
83
with 100 items per page.
- W4385800567 abstract "Modern Augmented reality applications require performing multiple tasks on each input frame simultaneously. Multi-task learning (MTL) represents an effective approach where multiple tasks share an encoder to extract representative features from the input frame, followed by task-specific decoders to generate predictions for each task. Generally, the shared encoder in MTL models needs to have a large representational capacity in order to generalize well to various tasks and input data, which has a negative effect on the inference latency. In this paper, we argue that due to the large variations in the complexity of the input frames, some computations might be unnecessary for the output. Therefore, we introduce AdaMTL, an adaptive framework that learns task-aware inference policies for the MTL models in an input-dependent manner Specifically, we attach a task-aware lightweight policy network to the shared encoder and co-train it alongside the MTL model to recognize unnecessary computations. During runtime, our task-aware policy network decides which parts of the model to activate depending on the input frame and the target computational complexity. Extensive experiments on the PASCAL dataset demonstrate that AdaMTL reduces the computational complexity by 43% while improving the accuracy by 1.32% compared to single-task models. Combined with SOTA MTL methodologies, AdaMTL boosts the accuracy by 7.8% while improving the efficiency by 3.1 ×. When deployed on Vuzix M4000 smart glasses, AdaMTL reduces the inference latency and the energy consumption by up to 21.8% and 37.5%, respectively, compared to the static MTL model. Our code is publicly available <sup xmlns:mml=http://www.w3.org/1998/Math/MathML xmlns:xlink=http://www.w3.org/1999/xlink>1</sup> ." @default.
- W4385800567 created "2023-08-15" @default.
- W4385800567 creator A5015719218 @default.
- W4385800567 creator A5043678248 @default.
- W4385800567 creator A5072797180 @default.
- W4385800567 date "2023-06-01" @default.
- W4385800567 modified "2023-09-26" @default.
- W4385800567 title "AdaMTL: Adaptive Input-dependent Inference for Efficient Multi-Task Learning" @default.
- W4385800567 cites W2031489346 @default.
- W4385800567 cites W2194775991 @default.
- W4385800567 cites W2798441115 @default.
- W4385800567 cites W2962864421 @default.
- W4385800567 cites W2963430933 @default.
- W4385800567 cites W2963877604 @default.
- W4385800567 cites W2964185501 @default.
- W4385800567 cites W3097777922 @default.
- W4385800567 cites W3138516171 @default.
- W4385800567 cites W3174295900 @default.
- W4385800567 cites W3193677416 @default.
- W4385800567 cites W3208912750 @default.
- W4385800567 cites W4214588794 @default.
- W4385800567 cites W4214893857 @default.
- W4385800567 cites W4221152259 @default.
- W4385800567 cites W4225930227 @default.
- W4385800567 cites W4312290555 @default.
- W4385800567 doi "https://doi.org/10.1109/cvprw59228.2023.00501" @default.
- W4385800567 hasPublicationYear "2023" @default.
- W4385800567 type Work @default.
- W4385800567 citedByCount "0" @default.
- W4385800567 crossrefType "proceedings-article" @default.
- W4385800567 hasAuthorship W4385800567A5015719218 @default.
- W4385800567 hasAuthorship W4385800567A5043678248 @default.
- W4385800567 hasAuthorship W4385800567A5072797180 @default.
- W4385800567 hasBestOaLocation W43858005672 @default.
- W4385800567 hasConcept C111919701 @default.
- W4385800567 hasConcept C11413529 @default.
- W4385800567 hasConcept C118505674 @default.
- W4385800567 hasConcept C119857082 @default.
- W4385800567 hasConcept C126042441 @default.
- W4385800567 hasConcept C154945302 @default.
- W4385800567 hasConcept C162324750 @default.
- W4385800567 hasConcept C175154964 @default.
- W4385800567 hasConcept C179799912 @default.
- W4385800567 hasConcept C187736073 @default.
- W4385800567 hasConcept C2776214188 @default.
- W4385800567 hasConcept C2780451532 @default.
- W4385800567 hasConcept C41008148 @default.
- W4385800567 hasConcept C45374587 @default.
- W4385800567 hasConcept C76155785 @default.
- W4385800567 hasConcept C82876162 @default.
- W4385800567 hasConceptScore W4385800567C111919701 @default.
- W4385800567 hasConceptScore W4385800567C11413529 @default.
- W4385800567 hasConceptScore W4385800567C118505674 @default.
- W4385800567 hasConceptScore W4385800567C119857082 @default.
- W4385800567 hasConceptScore W4385800567C126042441 @default.
- W4385800567 hasConceptScore W4385800567C154945302 @default.
- W4385800567 hasConceptScore W4385800567C162324750 @default.
- W4385800567 hasConceptScore W4385800567C175154964 @default.
- W4385800567 hasConceptScore W4385800567C179799912 @default.
- W4385800567 hasConceptScore W4385800567C187736073 @default.
- W4385800567 hasConceptScore W4385800567C2776214188 @default.
- W4385800567 hasConceptScore W4385800567C2780451532 @default.
- W4385800567 hasConceptScore W4385800567C41008148 @default.
- W4385800567 hasConceptScore W4385800567C45374587 @default.
- W4385800567 hasConceptScore W4385800567C76155785 @default.
- W4385800567 hasConceptScore W4385800567C82876162 @default.
- W4385800567 hasLocation W43858005671 @default.
- W4385800567 hasLocation W43858005672 @default.
- W4385800567 hasOpenAccess W4385800567 @default.
- W4385800567 hasPrimaryLocation W43858005671 @default.
- W4385800567 hasRelatedWork W2059508535 @default.
- W4385800567 hasRelatedWork W2082251630 @default.
- W4385800567 hasRelatedWork W2275988210 @default.
- W4385800567 hasRelatedWork W2946233749 @default.
- W4385800567 hasRelatedWork W2954183354 @default.
- W4385800567 hasRelatedWork W3030300968 @default.
- W4385800567 hasRelatedWork W3034672970 @default.
- W4385800567 hasRelatedWork W3159171600 @default.
- W4385800567 hasRelatedWork W3182043961 @default.
- W4385800567 hasRelatedWork W4285338583 @default.
- W4385800567 isParatext "false" @default.
- W4385800567 isRetracted "false" @default.
- W4385800567 workType "article" @default.