Matches in SemOpenAlex for { <https://semopenalex.org/work/W4319301108> ?p ?o ?g. }
- W4319301108 abstract "Multi-task learning promises better model generalization on a target task by jointly optimizing it with an auxiliary task. However, the current practice requires additional labeling efforts for the auxiliary task, while not guaranteeing better model performance. In this paper, we find that jointly training a dense prediction (target) task with a self-supervised (auxiliary) task can consistently improve the performance of the target task, while eliminating the need for labeling auxiliary tasks. We refer to this joint training as Composite Learning (CompL). Experiments of CompL on monocular depth estimation, semantic segmentation, and boundary detection show consistent performance improvements in fully and partially labeled datasets. Further analysis on depth estimation reveals that joint training with self-supervision outperforms most labeled auxiliary tasks. We also find that CompL can improve model robustness when the models are evaluated in new domains. These results demonstrate the benefits of self-supervision as an auxiliary task, and establish the design of novel task-specific self-supervised methods as a new axis of investigation for future multi-task learning research." @default.
- W4319301108 created "2023-02-07" @default.
- W4319301108 creator A5001254143 @default.
- W4319301108 creator A5051734787 @default.
- W4319301108 creator A5067487326 @default.
- W4319301108 creator A5079113064 @default.
- W4319301108 creator A5088685794 @default.
- W4319301108 date "2023-01-01" @default.
- W4319301108 modified "2023-09-27" @default.
- W4319301108 title "Composite Learning for Robust and Effective Dense Predictions" @default.
- W4319301108 cites W1555385401 @default.
- W4319301108 cites W1903029394 @default.
- W4319301108 cites W2031489346 @default.
- W4319301108 cites W2108598243 @default.
- W4319301108 cites W2110158442 @default.
- W4319301108 cites W2119823327 @default.
- W4319301108 cites W2144794286 @default.
- W4319301108 cites W2194775991 @default.
- W4319301108 cites W2216125271 @default.
- W4319301108 cites W2412782625 @default.
- W4319301108 cites W2913340405 @default.
- W4319301108 cites W2948647700 @default.
- W4319301108 cites W2962850830 @default.
- W4319301108 cites W2962864421 @default.
- W4319301108 cites W2963377935 @default.
- W4319301108 cites W2963488291 @default.
- W4319301108 cites W2963498646 @default.
- W4319301108 cites W2963881378 @default.
- W4319301108 cites W2964185501 @default.
- W4319301108 cites W2981952041 @default.
- W4319301108 cites W2991391304 @default.
- W4319301108 cites W2994633389 @default.
- W4319301108 cites W3035164673 @default.
- W4319301108 cites W3035524453 @default.
- W4319301108 cites W3035564946 @default.
- W4319301108 cites W3104240813 @default.
- W4319301108 cites W3122412340 @default.
- W4319301108 cites W3154799548 @default.
- W4319301108 cites W3165926952 @default.
- W4319301108 cites W3168822201 @default.
- W4319301108 cites W3168984673 @default.
- W4319301108 cites W3180060752 @default.
- W4319301108 cites W3180638445 @default.
- W4319301108 cites W3204397973 @default.
- W4319301108 cites W3204786153 @default.
- W4319301108 cites W343636949 @default.
- W4319301108 cites W4214745634 @default.
- W4319301108 cites W4221166941 @default.
- W4319301108 cites W4250482878 @default.
- W4319301108 cites W4285102522 @default.
- W4319301108 cites W845365781 @default.
- W4319301108 doi "https://doi.org/10.1109/wacv56688.2023.00233" @default.
- W4319301108 hasPublicationYear "2023" @default.
- W4319301108 type Work @default.
- W4319301108 citedByCount "0" @default.
- W4319301108 crossrefType "proceedings-article" @default.
- W4319301108 hasAuthorship W4319301108A5001254143 @default.
- W4319301108 hasAuthorship W4319301108A5051734787 @default.
- W4319301108 hasAuthorship W4319301108A5067487326 @default.
- W4319301108 hasAuthorship W4319301108A5079113064 @default.
- W4319301108 hasAuthorship W4319301108A5088685794 @default.
- W4319301108 hasBestOaLocation W43193011082 @default.
- W4319301108 hasConcept C104317684 @default.
- W4319301108 hasConcept C119857082 @default.
- W4319301108 hasConcept C127413603 @default.
- W4319301108 hasConcept C134306372 @default.
- W4319301108 hasConcept C154945302 @default.
- W4319301108 hasConcept C175154964 @default.
- W4319301108 hasConcept C177148314 @default.
- W4319301108 hasConcept C185592680 @default.
- W4319301108 hasConcept C201995342 @default.
- W4319301108 hasConcept C2776145971 @default.
- W4319301108 hasConcept C2780451532 @default.
- W4319301108 hasConcept C28006648 @default.
- W4319301108 hasConcept C33923547 @default.
- W4319301108 hasConcept C41008148 @default.
- W4319301108 hasConcept C51632099 @default.
- W4319301108 hasConcept C55493867 @default.
- W4319301108 hasConcept C63479239 @default.
- W4319301108 hasConcept C89600930 @default.
- W4319301108 hasConceptScore W4319301108C104317684 @default.
- W4319301108 hasConceptScore W4319301108C119857082 @default.
- W4319301108 hasConceptScore W4319301108C127413603 @default.
- W4319301108 hasConceptScore W4319301108C134306372 @default.
- W4319301108 hasConceptScore W4319301108C154945302 @default.
- W4319301108 hasConceptScore W4319301108C175154964 @default.
- W4319301108 hasConceptScore W4319301108C177148314 @default.
- W4319301108 hasConceptScore W4319301108C185592680 @default.
- W4319301108 hasConceptScore W4319301108C201995342 @default.
- W4319301108 hasConceptScore W4319301108C2776145971 @default.
- W4319301108 hasConceptScore W4319301108C2780451532 @default.
- W4319301108 hasConceptScore W4319301108C28006648 @default.
- W4319301108 hasConceptScore W4319301108C33923547 @default.
- W4319301108 hasConceptScore W4319301108C41008148 @default.
- W4319301108 hasConceptScore W4319301108C51632099 @default.
- W4319301108 hasConceptScore W4319301108C55493867 @default.
- W4319301108 hasConceptScore W4319301108C63479239 @default.
- W4319301108 hasConceptScore W4319301108C89600930 @default.
- W4319301108 hasLocation W43193011081 @default.
- W4319301108 hasLocation W43193011082 @default.