WALLACE: Weakly Supervised Learning of Deep Convolutional Neural Networks With Multiscale Evidence
This paper presents WALLACE, a new framework of deep convolutional neural networks, which perform ConvNet's pyramidal feature hierarchy for weakly supervised learning. Most prior works rely on the image pyramid or network ensemble, which is both complicated and usually expensive. Instead, WALLA...
| Main Authors: | , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
IEEE
2020-01-01
|
| Series: | IEEE Access |
| Subjects: | |
| Online Access: | https://ieeexplore.ieee.org/document/8966337/ |
| id |
doaj-5f607f89558b4abe9601c70950f71361 |
|---|---|
| record_format |
Article |
| spelling |
doaj-5f607f89558b4abe9601c70950f713612021-03-30T01:10:20ZengIEEEIEEE Access2169-35362020-01-018204492045810.1109/ACCESS.2020.29685458966337WALLACE: Weakly Supervised Learning of Deep Convolutional Neural Networks With Multiscale EvidenceYongsheng Liu0Wenyu Chen1Hong Qu2https://orcid.org/0000-0002-7590-144XTianlei Wang3Jiangzhou Ji4Kebin Miao5School of Computer Science and Engineering University of Electronic Science and Technology of China, Chengdu, ChinaSchool of Computer Science and Engineering University of Electronic Science and Technology of China, Chengdu, ChinaSchool of Computer Science and Engineering University of Electronic Science and Technology of China, Chengdu, ChinaSchool of Computer Science and Engineering University of Electronic Science and Technology of China, Chengdu, ChinaChina Merchants Bank AI Lab, Chengdu, ChinaChina Coal Research Institute, Beijing, ChinaThis paper presents WALLACE, a new framework of deep convolutional neural networks, which perform ConvNet's pyramidal feature hierarchy for weakly supervised learning. Most prior works rely on the image pyramid or network ensemble, which is both complicated and usually expensive. Instead, WALLACE is a more simple single-stage network that can predict objects present and location in an image without multiple rescale. Our model is trained efficiently using only global image-level labels, and it could generate meaningful multi-scale semantic feature maps by only one evaluation. Furthermore, a novel constrain-to-highlight loss is proposed to balances region selection among hierarchical feature maps, which additional improve model performance. Extensive experiments on object classification and weakly supervised pointwise object localization show that WALLACE achieves state-of-the-art results on the VOC 2007 and VOC 2012 benchmark without bells and whistles.https://ieeexplore.ieee.org/document/8966337/Weakly supervised learningconvolutional neural networksobject localizationobject classificationmulti-scale features |
| collection |
DOAJ |
| language |
English |
| format |
Article |
| sources |
DOAJ |
| author |
Yongsheng Liu Wenyu Chen Hong Qu Tianlei Wang Jiangzhou Ji Kebin Miao |
| spellingShingle |
Yongsheng Liu Wenyu Chen Hong Qu Tianlei Wang Jiangzhou Ji Kebin Miao WALLACE: Weakly Supervised Learning of Deep Convolutional Neural Networks With Multiscale Evidence IEEE Access Weakly supervised learning convolutional neural networks object localization object classification multi-scale features |
| author_facet |
Yongsheng Liu Wenyu Chen Hong Qu Tianlei Wang Jiangzhou Ji Kebin Miao |
| author_sort |
Yongsheng Liu |
| title |
WALLACE: Weakly Supervised Learning of Deep Convolutional Neural Networks With Multiscale Evidence |
| title_short |
WALLACE: Weakly Supervised Learning of Deep Convolutional Neural Networks With Multiscale Evidence |
| title_full |
WALLACE: Weakly Supervised Learning of Deep Convolutional Neural Networks With Multiscale Evidence |
| title_fullStr |
WALLACE: Weakly Supervised Learning of Deep Convolutional Neural Networks With Multiscale Evidence |
| title_full_unstemmed |
WALLACE: Weakly Supervised Learning of Deep Convolutional Neural Networks With Multiscale Evidence |
| title_sort |
wallace: weakly supervised learning of deep convolutional neural networks with multiscale evidence |
| publisher |
IEEE |
| series |
IEEE Access |
| issn |
2169-3536 |
| publishDate |
2020-01-01 |
| description |
This paper presents WALLACE, a new framework of deep convolutional neural networks, which perform ConvNet's pyramidal feature hierarchy for weakly supervised learning. Most prior works rely on the image pyramid or network ensemble, which is both complicated and usually expensive. Instead, WALLACE is a more simple single-stage network that can predict objects present and location in an image without multiple rescale. Our model is trained efficiently using only global image-level labels, and it could generate meaningful multi-scale semantic feature maps by only one evaluation. Furthermore, a novel constrain-to-highlight loss is proposed to balances region selection among hierarchical feature maps, which additional improve model performance. Extensive experiments on object classification and weakly supervised pointwise object localization show that WALLACE achieves state-of-the-art results on the VOC 2007 and VOC 2012 benchmark without bells and whistles. |
| topic |
Weakly supervised learning convolutional neural networks object localization object classification multi-scale features |
| url |
https://ieeexplore.ieee.org/document/8966337/ |
| work_keys_str_mv |
AT yongshengliu wallaceweaklysupervisedlearningofdeepconvolutionalneuralnetworkswithmultiscaleevidence AT wenyuchen wallaceweaklysupervisedlearningofdeepconvolutionalneuralnetworkswithmultiscaleevidence AT hongqu wallaceweaklysupervisedlearningofdeepconvolutionalneuralnetworkswithmultiscaleevidence AT tianleiwang wallaceweaklysupervisedlearningofdeepconvolutionalneuralnetworkswithmultiscaleevidence AT jiangzhouji wallaceweaklysupervisedlearningofdeepconvolutionalneuralnetworkswithmultiscaleevidence AT kebinmiao wallaceweaklysupervisedlearningofdeepconvolutionalneuralnetworkswithmultiscaleevidence |
| _version_ |
1724187587891953664 |