FRINK Linked Data Fragments server

Matches in SemOpenAlex for { <https://semopenalex.org/work/W3123825612> ?p ?o ?g. }

• W3123825612 endingPage "412" @default.
• W3123825612 startingPage "403" @default.
• W3123825612 abstract "Fine-grained image categorization is a challenging task due to the difficulty of localizing the discriminative regions for different sub-categories. Previous works mainly focus on using the manual annotations or the attention algorithm to localize these regions, which is demanding and complex in practical applications. This paper proposes a method of using a multi-level attention model (MLA-CNN) which has been trained on the full-size image train set of current tasks to localize the most discriminative regions. Intuitively, three typical receptive field sizes are selected for the multi-level attention maps. Then, multi-level dictionary learning is introduced to extract discriminative features from these localized regions. Our method explores a new thought about how to use the neural activations to generate multi-scale regions which are helpful for the fine-grained categorization. The method can be achieved in two steps. The first step is to select the neurons that have the max activation in the selected three feature maps. These feature maps are the outputs of the pre-trained CNN model by feeding the full-size images into the model. Then, we generate the discriminative regions according to the receptive field size of the selected neurons. The second step is to train the subtle networks with these multi-scale regions. One scaled discriminative region can be regarded as one typical dictionary feature. Then these results are integrated for final prediction. We evaluate our method on three challenging fine-grained image datasets, CUB-200-2011, Stanford Dogs, and Stanford Cars. The experimental results demonstrate that our method outperforms many state-of-the-art methods, using extra object/parts annotations and attention-based methods." @default.
• W3123825612 created "2021-02-01" @default.
• W3123825612 creator A5002896929 @default.
• W3123825612 creator A5023358501 @default.
• W3123825612 creator A5031803321 @default.
• W3123825612 creator A5064423286 @default.
• W3123825612 creator A5081796777 @default.
• W3123825612 date "2021-09-01" @default.
• W3123825612 modified "2023-09-29" @default.
• W3123825612 title "Multi-level dictionary learning for fine-grained images categorization with attention model" @default.
• W3123825612 cites W1432090866 @default.
• W3123825612 cites W1898560071 @default.
• W3123825612 cites W2049632392 @default.
• W3123825612 cites W2068562306 @default.
• W3123825612 cites W2159797991 @default.
• W3123825612 cites W2258484932 @default.
• W3123825612 cites W2289708887 @default.
• W3123825612 cites W2510373542 @default.
• W3123825612 cites W2599076810 @default.
• W3123825612 cites W2620694480 @default.
• W3123825612 cites W2625022521 @default.
• W3123825612 cites W2728364730 @default.
• W3123825612 cites W2731821979 @default.
• W3123825612 cites W2754803257 @default.
• W3123825612 cites W2763070548 @default.
• W3123825612 cites W2765995039 @default.
• W3123825612 cites W2804743778 @default.
• W3123825612 cites W2966167131 @default.
• W3123825612 cites W2972392244 @default.
• W3123825612 cites W2999458807 @default.
• W3123825612 cites W3000480515 @default.
• W3123825612 cites W3122238731 @default.
• W3123825612 cites W3124951096 @default.
• W3123825612 doi "https://doi.org/10.1016/j.neucom.2020.07.147" @default.
• W3123825612 hasPublicationYear "2021" @default.
• W3123825612 type Work @default.
• W3123825612 sameAs 3123825612 @default.
• W3123825612 citedByCount "3" @default.
• W3123825612 countsByYear W31238256122022 @default.
• W3123825612 countsByYear W31238256122023 @default.
• W3123825612 crossrefType "journal-article" @default.
• W3123825612 hasAuthorship W3123825612A5002896929 @default.
• W3123825612 hasAuthorship W3123825612A5023358501 @default.
• W3123825612 hasAuthorship W3123825612A5031803321 @default.
• W3123825612 hasAuthorship W3123825612A5064423286 @default.
• W3123825612 hasAuthorship W3123825612A5081796777 @default.
• W3123825612 hasConcept C119857082 @default.
• W3123825612 hasConcept C120665830 @default.
• W3123825612 hasConcept C121332964 @default.
• W3123825612 hasConcept C138885662 @default.
• W3123825612 hasConcept C153180895 @default.
• W3123825612 hasConcept C154945302 @default.
• W3123825612 hasConcept C162324750 @default.
• W3123825612 hasConcept C177264268 @default.
• W3123825612 hasConcept C187736073 @default.
• W3123825612 hasConcept C192209626 @default.
• W3123825612 hasConcept C199360897 @default.
• W3123825612 hasConcept C202444582 @default.
• W3123825612 hasConcept C2776401178 @default.
• W3123825612 hasConcept C2780451532 @default.
• W3123825612 hasConcept C33923547 @default.
• W3123825612 hasConcept C41008148 @default.
• W3123825612 hasConcept C41895202 @default.
• W3123825612 hasConcept C94124525 @default.
• W3123825612 hasConcept C9652623 @default.
• W3123825612 hasConcept C97931131 @default.
• W3123825612 hasConceptScore W3123825612C119857082 @default.
• W3123825612 hasConceptScore W3123825612C120665830 @default.
• W3123825612 hasConceptScore W3123825612C121332964 @default.
• W3123825612 hasConceptScore W3123825612C138885662 @default.
• W3123825612 hasConceptScore W3123825612C153180895 @default.
• W3123825612 hasConceptScore W3123825612C154945302 @default.
• W3123825612 hasConceptScore W3123825612C162324750 @default.
• W3123825612 hasConceptScore W3123825612C177264268 @default.
• W3123825612 hasConceptScore W3123825612C187736073 @default.
• W3123825612 hasConceptScore W3123825612C192209626 @default.
• W3123825612 hasConceptScore W3123825612C199360897 @default.
• W3123825612 hasConceptScore W3123825612C202444582 @default.
• W3123825612 hasConceptScore W3123825612C2776401178 @default.
• W3123825612 hasConceptScore W3123825612C2780451532 @default.
• W3123825612 hasConceptScore W3123825612C33923547 @default.
• W3123825612 hasConceptScore W3123825612C41008148 @default.
• W3123825612 hasConceptScore W3123825612C41895202 @default.
• W3123825612 hasConceptScore W3123825612C94124525 @default.
• W3123825612 hasConceptScore W3123825612C9652623 @default.
• W3123825612 hasConceptScore W3123825612C97931131 @default.
• W3123825612 hasFunder F4320321001 @default.
• W3123825612 hasLocation W31238256121 @default.
• W3123825612 hasOpenAccess W3123825612 @default.
• W3123825612 hasPrimaryLocation W31238256121 @default.
• W3123825612 hasRelatedWork W1652783584 @default.
• W3123825612 hasRelatedWork W1971623867 @default.
• W3123825612 hasRelatedWork W1982770690 @default.
• W3123825612 hasRelatedWork W1990254706 @default.
• W3123825612 hasRelatedWork W2024160000 @default.
• W3123825612 hasRelatedWork W2404514746 @default.
• W3123825612 hasRelatedWork W2743258233 @default.
• W3123825612 hasRelatedWork W2773500201 @default.

• next