1. Identificação | |
Tipo de Referência | Artigo em Revista Científica (Journal Article) |
Site | mtc-m21d.sid.inpe.br |
Código do Detentor | isadg {BR SPINPE} ibi 8JMKD3MGPCW/3DT298S |
Identificador | 8JMKD3MGP3W34T/47S47L8 |
Repositório | sid.inpe.br/mtc-m21d/2022/10.24.13.40 (acesso restrito) |
Última Atualização | 2022:10.24.13.40.37 (UTC) simone |
Repositório de Metadados | sid.inpe.br/mtc-m21d/2022/10.24.13.40.37 |
Última Atualização dos Metadados | 2023:01.03.16.46.21 (UTC) administrator |
DOI | 10.3389/fenvs.2022.946729 |
ISSN | 2296-665X |
Chave de Citação | WagnerSSHFLMYS:2022:KtSeHa |
Título | K-textures, a self-supervised hard clustering deep learning algorithm for satellite image segmentation |
Ano | 2022 |
Mês | Sept. |
Data de Acesso | 11 maio 2024 |
Tipo de Trabalho | journal article |
Tipo Secundário | PRE PI |
Número de Arquivos | 1 |
Tamanho | 9726 KiB |
|
2. Contextualização | |
Autor | 1 Wagner, Fabien Hubert 2 Silva, Ricardo Dalagnol 3 Sánchez Ipia, Alber Hamersson 4 Hirye, Mayumi C. M. 5 Favrichon, Samuel 6 Lee, Jake H. 7 Mauceri, Steffen 8 Yang, Yan 9 Saatchi, Sassan |
Grupo | 1 YYY-CGCT-INPE-MCTI-GOV-BR 2 SER-SRE-DIPGR-INPE-MCTI-GOV-BR 3 YYY-CGCT-INPE-MCTI-GOV-BR |
Afiliação | 1 Instituto Nacional de Pesquisas Espaciais (INPE) 2 Instituto Nacional de Pesquisas Espaciais (INPE) 3 Instituto Nacional de Pesquisas Espaciais (INPE) 4 Universidade de São Paulo (USP) 5 NASA-Jet Propulsion Laboratory 6 NASA-Jet Propulsion Laboratory 7 NASA-Jet Propulsion Laboratory 8 NASA-Jet Propulsion Laboratory 9 NASA-Jet Propulsion Laboratory |
Endereço de e-Mail do Autor | 1 wagner.h.fabien@gmail.com 2 ricds@hotmail.com 3 albhasan@gmail.com |
Revista | Frontiers in Environmental Science |
Volume | 10 |
Páginas | e946729 |
Histórico (UTC) | 2022-10-24 13:42:21 :: simone -> administrator :: 2022 2023-01-03 16:46:21 :: administrator -> simone :: 2022 |
|
3. Conteúdo e estrutura | |
É a matriz ou uma cópia? | é a matriz |
Estágio do Conteúdo | concluido |
Transferível | 1 |
Tipo do Conteúdo | External Contribution |
Tipo de Versão | publisher |
Palavras-Chave | deep learning - artificial neural network discrete optimization algorithm landcover planetscope satellite segmentation (image processing) self-supervised segmentation tensorflow (2) tropical forest |
Resumo | Deep learning self-supervised algorithms that can segment an image in a fixed number of hard clusters such as the k-means algorithm and with an end-to-end deep learning approach are still lacking. Here, we introduce the k-textures algorithm which provides self-supervised segmentation of a 4-band image (RGB-NIR) for a k number of classes. An example of its application on high-resolution Planet satellite imagery is given. Our algorithm shows that discrete search is feasible using convolutional neural networks (CNN) and gradient descent. The model detects k hard clustering classes represented in the model as k discrete binary masks and their associated k independently generated textures, which combined are a simulation of the original image. The similarity loss is the mean squared error between the features of the original and the simulated image, both extracted from the penultimate convolutional block of Keras imagenet pre-trained VGG-16 model and a custom feature extractor made with Planet data. The main advances of the k-textures model are: first, the k discrete binary masks are obtained inside the model using gradient descent. The model allows for the generation of discrete binary masks using a novel method using a hard sigmoid activation function. Second, it provides hard clustering classeseach pixel has only one class. Finally, in comparison to k-means, where each pixel is considered independently, here, contextual information is also considered and each class is not associated only with similar values in the color channels but with a texture. Our approach is designed to ease the production of training samples for satellite image segmentation and the k-textures architecture could be adapted to support different numbers of bands and for more complex self-segmentation tasks, such as object self-segmentation. The model codes and weights are available at https://doi.org/10.5281/zenodo.6359859. |
Área | SRE |
Arranjo 1 | urlib.net > BDMCI > Fonds > Produção pgr ATUAIS > SER > K-textures, a self-supervised... |
Arranjo 2 | urlib.net > BDMCI > Fonds > Produção a partir de 2021 > CGCT > K-textures, a self-supervised... |
Conteúdo da Pasta doc | acessar |
Conteúdo da Pasta source | não têm arquivos |
Conteúdo da Pasta agreement | |
|
4. Condições de acesso e uso | |
Idioma | en |
Arquivo Alvo | fenvs-10-946729.pdf |
Grupo de Usuários | simone |
Grupo de Leitores | administrator simone |
Visibilidade | shown |
Permissão de Leitura | deny from all and allow from 150.163 |
Permissão de Atualização | não transferida |
|
5. Fontes relacionadas | |
Unidades Imediatamente Superiores | 8JMKD3MGPCW/3F3NU5S 8JMKD3MGPCW/46KUATE |
Lista de Itens Citando | sid.inpe.br/bibdigital/2013/10.18.22.34 3 sid.inpe.br/bibdigital/2022/04.03.22.23 1 |
Divulgação | PORTALCAPES |
Acervo Hospedeiro | urlib.net/www/2021/06.04.03.40 |
|
6. Notas | |
Campos Vazios | alternatejournal archivingpolicy archivist callnumber copyholder copyright creatorhistory descriptionlevel e-mailaddress format isbn label lineage mark mirrorrepository nextedition notes number orcid parameterlist parentrepositories previousedition previouslowerunit progress project resumeid rightsholder schedulinginformation secondarydate secondarykey secondarymark session shorttitle sponsor subject tertiarymark tertiarytype url |
|
7. Controle da descrição | |
e-Mail (login) | simone |
atualizar | |
|