The original paper is in English. Non-English content has been machine-translated and may contain typographical errors or mistranslations. ex. Some numerals are expressed as "XNUMX".
Copyrights notice
The original paper is in English. Non-English content has been machine-translated and may contain typographical errors or mistranslations. Copyrights notice
A localização e a representação de recursos das partes do objeto desempenham papéis importantes no reconhecimento visual refinado. Para promover a precisão do reconhecimento final sem caixas delimitadoras/anotações de peças, muitos estudos adotam redes de localização de objetos para propor caixas delimitadoras/anotações de peças apenas com rótulos de categoria e, em seguida, cortam as imagens em imagens parciais para ajudar a rede de classificação a tomar a decisão final. Em nosso trabalho, para propor imagens parciais mais informativas e extrair efetivamente características discriminativas das imagens originais e parciais, propomos uma abordagem em duas etapas que pode fundir as características originais e parciais, avaliando e classificando as informações das imagens parciais. Os resultados experimentais mostram que a abordagem proposta atinge um excelente desempenho em dois conjuntos de dados de referência, o que demonstra a sua eficácia.
Kangbo SUN
Shanghai Jiao Tong University
Jie ZHU
Shanghai Jiao Tong University
The copyright of the original papers published on this site belongs to IEICE. Unauthorized use of the original or translated papers is prohibited. See IEICE Provisions on Copyright for details.
Copiar
Kangbo SUN, Jie ZHU, "A Two-Stage Approach for Fine-Grained Visual Recognition via Confidence Ranking and Fusion" in IEICE TRANSACTIONS on Information,
vol. E103-D, no. 12, pp. 2693-2700, December 2020, doi: 10.1587/transinf.2020EDP7024.
Abstract: Location and feature representation of object's parts play key roles in fine-grained visual recognition. To promote the final recognition accuracy without any bounding boxes/part annotations, many studies adopt object location networks to propose bounding boxes/part annotations with only category labels, and then crop the images into partial images to help the classification network make the final decision. In our work, to propose more informative partial images and effectively extract discriminative features from the original and partial images, we propose a two-stage approach that can fuse the original features and partial features by evaluating and ranking the information of partial images. Experimental results show that our proposed approach achieves excellent performance on two benchmark datasets, which demonstrates its effectiveness.
URL: https://global.ieice.org/en_transactions/information/10.1587/transinf.2020EDP7024/_p
Copiar
@ARTICLE{e103-d_12_2693,
author={Kangbo SUN, Jie ZHU, },
journal={IEICE TRANSACTIONS on Information},
title={A Two-Stage Approach for Fine-Grained Visual Recognition via Confidence Ranking and Fusion},
year={2020},
volume={E103-D},
number={12},
pages={2693-2700},
abstract={Location and feature representation of object's parts play key roles in fine-grained visual recognition. To promote the final recognition accuracy without any bounding boxes/part annotations, many studies adopt object location networks to propose bounding boxes/part annotations with only category labels, and then crop the images into partial images to help the classification network make the final decision. In our work, to propose more informative partial images and effectively extract discriminative features from the original and partial images, we propose a two-stage approach that can fuse the original features and partial features by evaluating and ranking the information of partial images. Experimental results show that our proposed approach achieves excellent performance on two benchmark datasets, which demonstrates its effectiveness.},
keywords={},
doi={10.1587/transinf.2020EDP7024},
ISSN={1745-1361},
month={December},}
Copiar
TY - JOUR
TI - A Two-Stage Approach for Fine-Grained Visual Recognition via Confidence Ranking and Fusion
T2 - IEICE TRANSACTIONS on Information
SP - 2693
EP - 2700
AU - Kangbo SUN
AU - Jie ZHU
PY - 2020
DO - 10.1587/transinf.2020EDP7024
JO - IEICE TRANSACTIONS on Information
SN - 1745-1361
VL - E103-D
IS - 12
JA - IEICE TRANSACTIONS on Information
Y1 - December 2020
AB - Location and feature representation of object's parts play key roles in fine-grained visual recognition. To promote the final recognition accuracy without any bounding boxes/part annotations, many studies adopt object location networks to propose bounding boxes/part annotations with only category labels, and then crop the images into partial images to help the classification network make the final decision. In our work, to propose more informative partial images and effectively extract discriminative features from the original and partial images, we propose a two-stage approach that can fuse the original features and partial features by evaluating and ranking the information of partial images. Experimental results show that our proposed approach achieves excellent performance on two benchmark datasets, which demonstrates its effectiveness.
ER -