FRINK Linked Data Fragments server

Matches in SemOpenAlex for { <https://semopenalex.org/work/W3206354416> ?p ?o ?g. }

• W3206354416 abstract "Deep learning in the presence of noisy annotations has been studied extensively in classification, but much less in segmentation tasks. In this work, we study the learning dynamics of deep segmentation networks trained on inaccurately-annotated data. We discover a phenomenon that has been previously reported in the context of classification: the networks tend to first fit the clean pixel-level labels during an early-learning phase, before eventually memorizing the false annotations. However, in contrast to classification, memorization in segmentation does not arise simultaneously for all semantic categories. Inspired by these findings, we propose a new method for segmentation from noisy annotations with two key elements. First, we detect the beginning of the memorization phase separately for each category during training. This allows us to adaptively correct the noisy annotations in order to exploit early learning. Second, we incorporate a regularization term that enforces consistency across scales to boost robustness against annotation noise. Our method outperforms standard approaches on a medical-imaging segmentation task where noises are synthesized to mimic human annotation errors. It also provides robustness to realistic noisy annotations present in weakly-supervised semantic segmentation, achieving state-of-the-art results on PASCAL VOC 2012. Code is available at https://github.com/Kangningthu/ADELE" @default.
• W3206354416 created "2021-10-25" @default.
• W3206354416 creator A5001916762 @default.
• W3206354416 creator A5003735451 @default.
• W3206354416 creator A5034482431 @default.
• W3206354416 creator A5044336556 @default.
• W3206354416 creator A5066499149 @default.
• W3206354416 date "2021-10-07" @default.
• W3206354416 modified "2023-10-16" @default.
• W3206354416 title "Adaptive Early-Learning Correction for Segmentation from Noisy Annotations" @default.
• W3206354416 cites W1495267108 @default.
• W3206354416 cites W1884191083 @default.
• W3206354416 cites W1901129140 @default.
• W3206354416 cites W2037227137 @default.
• W3206354416 cites W2096579040 @default.
• W3206354416 cites W2100209835 @default.
• W3206354416 cites W2108598243 @default.
• W3206354416 cites W2133515615 @default.
• W3206354416 cites W2144794286 @default.
• W3206354416 cites W2161185676 @default.
• W3206354416 cites W2161236525 @default.
• W3206354416 cites W2295107390 @default.
• W3206354416 cites W2306289963 @default.
• W3206354416 cites W2337429362 @default.
• W3206354416 cites W2412782625 @default.
• W3206354416 cites W2558580397 @default.
• W3206354416 cites W2560023338 @default.
• W3206354416 cites W2592691248 @default.
• W3206354416 cites W2600144439 @default.
• W3206354416 cites W2797594039 @default.
• W3206354416 cites W2798376494 @default.
• W3206354416 cites W2799124825 @default.
• W3206354416 cites W2799213142 @default.
• W3206354416 cites W2886773299 @default.
• W3206354416 cites W2895420332 @default.
• W3206354416 cites W2904856451 @default.
• W3206354416 cites W2907965334 @default.
• W3206354416 cites W2956648669 @default.
• W3206354416 cites W2962758679 @default.
• W3206354416 cites W2962762541 @default.
• W3206354416 cites W2962867364 @default.
• W3206354416 cites W2963091558 @default.
• W3206354416 cites W2963096987 @default.
• W3206354416 cites W2963163009 @default.
• W3206354416 cites W2964159205 @default.
• W3206354416 cites W2964274690 @default.
• W3206354416 cites W2964288706 @default.
• W3206354416 cites W2967052791 @default.
• W3206354416 cites W2978426779 @default.
• W3206354416 cites W2979338721 @default.
• W3206354416 cites W2980061058 @default.
• W3206354416 cites W2982093251 @default.
• W3206354416 cites W2991083560 @default.
• W3206354416 cites W2996952120 @default.
• W3206354416 cites W2997664062 @default.
• W3206354416 cites W3001197829 @default.
• W3206354416 cites W3033272814 @default.
• W3206354416 cites W3034333089 @default.
• W3206354416 cites W3034373787 @default.
• W3206354416 cites W3034930876 @default.
• W3206354416 cites W3046830682 @default.
• W3206354416 cites W3098191145 @default.
• W3206354416 cites W3100040694 @default.
• W3206354416 cites W3102583815 @default.
• W3206354416 cites W3102714038 @default.
• W3206354416 cites W3107653507 @default.
• W3206354416 cites W3113231417 @default.
• W3206354416 cites W3123795136 @default.
• W3206354416 cites W3130880347 @default.
• W3206354416 cites W3147542689 @default.
• W3206354416 cites W3149621864 @default.
• W3206354416 doi "https://doi.org/10.48550/arxiv.2110.03740" @default.
• W3206354416 hasPublicationYear "2021" @default.
• W3206354416 type Work @default.
• W3206354416 sameAs 3206354416 @default.
• W3206354416 citedByCount "0" @default.
• W3206354416 crossrefType "posted-content" @default.
• W3206354416 hasAuthorship W3206354416A5001916762 @default.
• W3206354416 hasAuthorship W3206354416A5003735451 @default.
• W3206354416 hasAuthorship W3206354416A5034482431 @default.
• W3206354416 hasAuthorship W3206354416A5044336556 @default.
• W3206354416 hasAuthorship W3206354416A5066499149 @default.
• W3206354416 hasBestOaLocation W32063544161 @default.
• W3206354416 hasConcept C104317684 @default.
• W3206354416 hasConcept C108583219 @default.
• W3206354416 hasConcept C119857082 @default.
• W3206354416 hasConcept C145420912 @default.
• W3206354416 hasConcept C153180895 @default.
• W3206354416 hasConcept C154945302 @default.
• W3206354416 hasConcept C165696696 @default.
• W3206354416 hasConcept C185592680 @default.
• W3206354416 hasConcept C199360897 @default.
• W3206354416 hasConcept C2776321320 @default.
• W3206354416 hasConcept C30038468 @default.
• W3206354416 hasConcept C33923547 @default.
• W3206354416 hasConcept C38652104 @default.
• W3206354416 hasConcept C41008148 @default.
• W3206354416 hasConcept C55493867 @default.
• W3206354416 hasConcept C63479239 @default.
• W3206354416 hasConcept C75608658 @default.

• next