Matches in SemOpenAlex for { <https://semopenalex.org/work/W4296339905> ?p ?o ?g. }
Showing items 1 to 56 of
56
with 100 items per page.
- W4296339905 abstract "For precise pixel-level crack extraction, a large amount of densely labeled data is usually required to train a deep neural network, which is time-consuming and laborious. Scribble annotations are easier to label crack images, especially for tiny cracks. Due to the topology complexity, irregular edges, and sparse annotations, crack segmentation with scribbles is a challenging task. In this paper, we first relabel a crack dataset with scribbles. Inspired by the long-range dependencies of Vision Transformer, we propose a novel feature encoder with hybrid convolutional neural network and Vision Transformer to extract low-level features, high-level features and global context features. Moreover, a skip feature aggregation module is adopted to fuse the three types of features. Additionally, in order to alleviate the poor boundary localization caused by sparse annotations, we adopt the deep supervision strategy with two auxiliary losses to train the network. Extensive experiments on the crack dataset demonstrate that our method outperforms existing weakly-supervised methods and is on par with fully-supervised methods." @default.
- W4296339905 created "2022-09-20" @default.
- W4296339905 creator A5010677746 @default.
- W4296339905 creator A5015102287 @default.
- W4296339905 creator A5077138592 @default.
- W4296339905 date "2022-07-20" @default.
- W4296339905 modified "2023-09-24" @default.
- W4296339905 title "Weakly-Supervised Crack Segmentation via Scribble Annotations" @default.
- W4296339905 cites W1495267108 @default.
- W4296339905 cites W1903029394 @default.
- W4296339905 cites W1927251054 @default.
- W4296339905 cites W2033318022 @default.
- W4296339905 cites W2133665775 @default.
- W4296339905 cites W2194775991 @default.
- W4296339905 cites W2552414813 @default.
- W4296339905 cites W2598457882 @default.
- W4296339905 cites W2768955070 @default.
- W4296339905 cites W2899242765 @default.
- W4296339905 cites W2912350898 @default.
- W4296339905 cites W2962758679 @default.
- W4296339905 cites W2963020481 @default.
- W4296339905 cites W2964308596 @default.
- W4296339905 cites W2997316506 @default.
- W4296339905 cites W3011233732 @default.
- W4296339905 cites W3034453930 @default.
- W4296339905 cites W3084580419 @default.
- W4296339905 cites W3127550792 @default.
- W4296339905 cites W3177087374 @default.
- W4296339905 cites W3202439968 @default.
- W4296339905 cites W3202498229 @default.
- W4296339905 cites W4221161499 @default.
- W4296339905 doi "https://doi.org/10.1109/icsip55141.2022.9886030" @default.
- W4296339905 hasPublicationYear "2022" @default.
- W4296339905 type Work @default.
- W4296339905 citedByCount "0" @default.
- W4296339905 crossrefType "proceedings-article" @default.
- W4296339905 hasAuthorship W4296339905A5010677746 @default.
- W4296339905 hasAuthorship W4296339905A5015102287 @default.
- W4296339905 hasAuthorship W4296339905A5077138592 @default.
- W4296339905 hasFunder F4320321001 @default.
- W4296339905 hasLocation W42963399051 @default.
- W4296339905 hasOpenAccess W4296339905 @default.
- W4296339905 hasPrimaryLocation W42963399051 @default.
- W4296339905 hasRelatedWork W2279398222 @default.
- W4296339905 hasRelatedWork W2406522397 @default.
- W4296339905 hasRelatedWork W2732542196 @default.
- W4296339905 hasRelatedWork W2760085659 @default.
- W4296339905 hasRelatedWork W2773120646 @default.
- W4296339905 hasRelatedWork W2795329967 @default.
- W4296339905 hasRelatedWork W3102253946 @default.
- W4296339905 hasRelatedWork W3144574764 @default.
- W4296339905 hasRelatedWork W4226289457 @default.
- W4296339905 hasRelatedWork W4293211451 @default.
- W4296339905 isParatext "false" @default.
- W4296339905 isRetracted "false" @default.
- W4296339905 workType "article" @default.