FRINK Linked Data Fragments server

Matches in SemOpenAlex for { <https://semopenalex.org/work/W3216631677> ?p ?o ?g. }

• W3216631677 endingPage "9314" @default.
• W3216631677 startingPage "9298" @default.
• W3216631677 abstract "Transfer learning enables to re-use knowledge learned on a source task to help learning a target task. A simple form of transfer learning is common in current state-of-the-art computer vision models, i.e., pre-training a model for image classification on the ILSVRC dataset, and then fine-tune on any target task. However, previous systematic studies of transfer learning have been limited and the circumstances in which it is expected to work are not fully understood. In this paper we carry out an extensive experimental exploration of transfer learning across vastly different image domains (consumer photos, autonomous driving, aerial imagery, underwater, indoor scenes, synthetic, close-ups) and task types (semantic segmentation, object detection, depth estimation, keypoint detection). Importantly, these are all complex, structured output tasks types relevant to modern computer vision applications. In total we carry out over 2000 transfer learning experiments, including many where the source and target come from different image domains, task types, or both. We systematically analyze these experiments to understand the impact of image domain, task type, and dataset size on transfer learning performance. Our study leads to several insights and concrete recommendations: (1) for most tasks there exists a source which significantly outperforms ILSVRC’12 pre-training; (2) the image domain is the most important factor for achieving positive transfer; (3) the source dataset should <italic xmlns:mml=http://www.w3.org/1998/Math/MathML xmlns:xlink=http://www.w3.org/1999/xlink>include</i> the image domain of the target dataset to achieve best results; (4) at the same time, we observe only small negative effects when the image domain of the source task is much broader than that of the target; (5) transfer across task types can be beneficial, but its success is heavily dependent on both the source and target task types." @default.
• W3216631677 created "2021-12-06" @default.
• W3216631677 creator A5003940396 @default.
• W3216631677 creator A5020938231 @default.
• W3216631677 creator A5039948189 @default.
• W3216631677 creator A5056112540 @default.
• W3216631677 creator A5063267153 @default.
• W3216631677 date "2022-12-01" @default.
• W3216631677 modified "2023-10-18" @default.
• W3216631677 title "Factors of Influence for Transfer Learning Across Diverse Appearance Domains and Task Types" @default.
• W3216631677 cites W125693051 @default.
• W3216631677 cites W1536680647 @default.
• W3216631677 cites W1722318740 @default.
• W3216631677 cites W1923184257 @default.
• W3216631677 cites W1985238052 @default.
• W3216631677 cites W1985514943 @default.
• W3216631677 cites W1992178727 @default.
• W3216631677 cites W2015563892 @default.
• W3216631677 cites W2021259423 @default.
• W3216631677 cites W2037227137 @default.
• W3216631677 cites W2080873731 @default.
• W3216631677 cites W2098883970 @default.
• W3216631677 cites W2100031962 @default.
• W3216631677 cites W2108598243 @default.
• W3216631677 cites W2117539524 @default.
• W3216631677 cites W2125215748 @default.
• W3216631677 cites W2171943915 @default.
• W3216631677 cites W2194775991 @default.
• W3216631677 cites W2203224402 @default.
• W3216631677 cites W2307770531 @default.
• W3216631677 cites W2340897893 @default.
• W3216631677 cites W2560647685 @default.
• W3216631677 cites W2561196672 @default.
• W3216631677 cites W2575615142 @default.
• W3216631677 cites W2584009249 @default.
• W3216631677 cites W2591924527 @default.
• W3216631677 cites W2593768305 @default.
• W3216631677 cites W2594519801 @default.
• W3216631677 cites W2605488490 @default.
• W3216631677 cites W2617027347 @default.
• W3216631677 cites W2627183927 @default.
• W3216631677 cites W2737258237 @default.
• W3216631677 cites W2743627947 @default.
• W3216631677 cites W2781228439 @default.
• W3216631677 cites W2796346823 @default.
• W3216631677 cites W2798836702 @default.
• W3216631677 cites W2804935296 @default.
• W3216631677 cites W2885018852 @default.
• W3216631677 cites W2901870313 @default.
• W3216631677 cites W2948604360 @default.
• W3216631677 cites W2949736877 @default.
• W3216631677 cites W2949813473 @default.
• W3216631677 cites W2962685835 @default.
• W3216631677 cites W2962749812 @default.
• W3216631677 cites W2962843773 @default.
• W3216631677 cites W2962966271 @default.
• W3216631677 cites W2963078159 @default.
• W3216631677 cites W2963107255 @default.
• W3216631677 cites W2963150697 @default.
• W3216631677 cites W2963488291 @default.
• W3216631677 cites W2963703197 @default.
• W3216631677 cites W2963870446 @default.
• W3216631677 cites W2964105864 @default.
• W3216631677 cites W2964162504 @default.
• W3216631677 cites W2964185501 @default.
• W3216631677 cites W2964288524 @default.
• W3216631677 cites W2965954137 @default.
• W3216631677 cites W2982102242 @default.
• W3216631677 cites W2982453938 @default.
• W3216631677 cites W2986782335 @default.
• W3216631677 cites W2990230185 @default.
• W3216631677 cites W2999219213 @default.
• W3216631677 cites W3014641072 @default.
• W3216631677 cites W3034836583 @default.
• W3216631677 cites W3035086574 @default.
• W3216631677 cites W3035272520 @default.
• W3216631677 cites W3035564946 @default.
• W3216631677 cites W3097217077 @default.
• W3216631677 cites W3097651496 @default.
• W3216631677 cites W3108975329 @default.
• W3216631677 cites W3109055604 @default.
• W3216631677 cites W3109083691 @default.
• W3216631677 cites W3132413796 @default.
• W3216631677 cites W3173206925 @default.
• W3216631677 cites W3176276772 @default.
• W3216631677 cites W4288083516 @default.
• W3216631677 doi "https://doi.org/10.1109/tpami.2021.3129870" @default.
• W3216631677 hasPubMedId "https://pubmed.ncbi.nlm.nih.gov/34813469" @default.
• W3216631677 hasPublicationYear "2022" @default.
• W3216631677 type Work @default.
• W3216631677 sameAs 3216631677 @default.
• W3216631677 citedByCount "19" @default.
• W3216631677 countsByYear W32166316772021 @default.
• W3216631677 countsByYear W32166316772022 @default.
• W3216631677 countsByYear W32166316772023 @default.
• W3216631677 crossrefType "journal-article" @default.
• W3216631677 hasAuthorship W3216631677A5003940396 @default.
• W3216631677 hasAuthorship W3216631677A5020938231 @default.

• next