Matches in SemOpenAlex for { <https://semopenalex.org/work/W3205377930> ?p ?o ?g. }
- W3205377930 abstract "Weakly supervised semantic segmentation produces pixel-level localization from class labels; however, a classifier trained on such labels is likely to focus on a small discriminative region of the target object. We interpret this phenomenon using the information bottleneck principle: the final layer of a deep neural network, activated by the sigmoid or softmax activation functions, causes an information bottleneck, and as a result, only a subset of the task-relevant information is passed on to the output. We first support this argument through a simulated toy experiment and then propose a method to reduce the information bottleneck by removing the last activation function. In addition, we introduce a new pooling method that further encourages the transmission of information from non-discriminative regions to the classification. Our experimental evaluations demonstrate that this simple modification significantly improves the quality of localization maps on both the PASCAL VOC 2012 and MS COCO 2014 datasets, exhibiting a new state-of-the-art performance for weakly supervised semantic segmentation. The code is available at: this https URL" @default.
- W3205377930 created "2021-10-25" @default.
- W3205377930 creator A5001243382 @default.
- W3205377930 creator A5022314272 @default.
- W3205377930 creator A5086877012 @default.
- W3205377930 creator A5091880855 @default.
- W3205377930 date "2021-10-13" @default.
- W3205377930 modified "2023-09-27" @default.
- W3205377930 title "Reducing Information Bottleneck for Weakly Supervised Semantic Segmentation." @default.
- W3205377930 cites W1686946872 @default.
- W3205377930 cites W1861492603 @default.
- W3205377930 cites W1945608308 @default.
- W3205377930 cites W2031489346 @default.
- W3205377930 cites W2099111195 @default.
- W3205377930 cites W2099471712 @default.
- W3205377930 cites W2108598243 @default.
- W3205377930 cites W2144794286 @default.
- W3205377930 cites W2161236525 @default.
- W3205377930 cites W2194775991 @default.
- W3205377930 cites W2295107390 @default.
- W3205377930 cites W2306289963 @default.
- W3205377930 cites W2412782625 @default.
- W3205377930 cites W2593634001 @default.
- W3205377930 cites W2683470288 @default.
- W3205377930 cites W2750562598 @default.
- W3205377930 cites W2785885194 @default.
- W3205377930 cites W2798232928 @default.
- W3205377930 cites W2799124825 @default.
- W3205377930 cites W2805977310 @default.
- W3205377930 cites W2899771611 @default.
- W3205377930 cites W2940181122 @default.
- W3205377930 cites W2950157285 @default.
- W3205377930 cites W2950373644 @default.
- W3205377930 cites W2953139137 @default.
- W3205377930 cites W2954400107 @default.
- W3205377930 cites W2956648669 @default.
- W3205377930 cites W2962705976 @default.
- W3205377930 cites W2962758679 @default.
- W3205377930 cites W2962858109 @default.
- W3205377930 cites W2963198662 @default.
- W3205377930 cites W2963346885 @default.
- W3205377930 cites W2963446712 @default.
- W3205377930 cites W2963606198 @default.
- W3205377930 cites W2964184826 @default.
- W3205377930 cites W2967958187 @default.
- W3205377930 cites W2980189057 @default.
- W3205377930 cites W2981689412 @default.
- W3205377930 cites W2991083560 @default.
- W3205377930 cites W2994871715 @default.
- W3205377930 cites W2996952120 @default.
- W3205377930 cites W2997076984 @default.
- W3205377930 cites W2997851315 @default.
- W3205377930 cites W2999480731 @default.
- W3205377930 cites W3015272069 @default.
- W3205377930 cites W3032908737 @default.
- W3205377930 cites W3035703639 @default.
- W3205377930 cites W3047498618 @default.
- W3205377930 cites W3082755608 @default.
- W3205377930 cites W3085685449 @default.
- W3205377930 cites W3093151588 @default.
- W3205377930 cites W3098191145 @default.
- W3205377930 cites W3102659010 @default.
- W3205377930 cites W3102714038 @default.
- W3205377930 cites W3103376464 @default.
- W3205377930 cites W3103634170 @default.
- W3205377930 cites W3106349512 @default.
- W3205377930 cites W3107653507 @default.
- W3205377930 cites W3128082729 @default.
- W3205377930 cites W3132452637 @default.
- W3205377930 cites W3138863770 @default.
- W3205377930 cites W3147542689 @default.
- W3205377930 cites W3173957243 @default.
- W3205377930 cites W3177958285 @default.
- W3205377930 hasPublicationYear "2021" @default.
- W3205377930 type Work @default.
- W3205377930 sameAs 3205377930 @default.
- W3205377930 citedByCount "0" @default.
- W3205377930 crossrefType "posted-content" @default.
- W3205377930 hasAuthorship W3205377930A5001243382 @default.
- W3205377930 hasAuthorship W3205377930A5022314272 @default.
- W3205377930 hasAuthorship W3205377930A5086877012 @default.
- W3205377930 hasAuthorship W3205377930A5091880855 @default.
- W3205377930 hasConcept C119857082 @default.
- W3205377930 hasConcept C149635348 @default.
- W3205377930 hasConcept C153180895 @default.
- W3205377930 hasConcept C154945302 @default.
- W3205377930 hasConcept C188441871 @default.
- W3205377930 hasConcept C199360897 @default.
- W3205377930 hasConcept C2780513914 @default.
- W3205377930 hasConcept C41008148 @default.
- W3205377930 hasConcept C46686674 @default.
- W3205377930 hasConcept C50644808 @default.
- W3205377930 hasConcept C60008888 @default.
- W3205377930 hasConcept C70437156 @default.
- W3205377930 hasConcept C73555534 @default.
- W3205377930 hasConcept C75608658 @default.
- W3205377930 hasConcept C8521452 @default.
- W3205377930 hasConcept C89600930 @default.
- W3205377930 hasConcept C95623464 @default.
- W3205377930 hasConcept C97931131 @default.