MoE-SPNet: A mixture-of-experts scene parsing network

Fu, Huan; Gong, Maoguo; Wang, Chaohui; Tao, Dacheng

doi:10.1016/j.patcog.2018.07.020

Cited by 17 publications

(4 citation statements)

References 61 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Yuan et al [32] propose the pyramid object context to model the category dependencies. Fu et al [33] propose a mixture-of-experts based scene parsing network that incorporates a convolutional mixture-of-experts layer to assess the importance of features from different levels. Zhao et al [34] propose the point-wise spatial attention network to relax the local neighborhood constraint.…”

Section: B Fcn Based Methodsmentioning

confidence: 99%

“…FCN backbone Pixel Acc.% Mean IoU% FCN8s [4] VGG-16 71.32 29.39 FCN8s [4] ResNet-50 74.57 34.38 FCN8s [4] ResNet-101 74.83 34.62 DilatedNet [5] VGG-16 73.55 32.31 SegNet [7] VGG-16 71.00 21.64 CascadeNet [3] VGG-16 74.52 34.90 RefineNet [46] ResNet-101 -40.20 RefineNet [46] ResNet-152 -40.71 PSPNet [8] ResNet-50 80.04 41.68 PSPNet [8] ResNet-101 81.39 43.29 PSPNet [8] ResNet-269 81.69 44.94 SAC [48] ResNet-101 81.86 44.30 DeepLabv2 [9] VGG-16 74.88 33.03 DeepLabv2 [9] ResNet-101 77.31 36.75 MoE-SPNet [33] VGG-16 75.50 34.35 MoE-SPNet [33] ResNet-101 78.02 37.89 EncNet [27] ResNet-50 79.73 41.11 EncNet [27] ResNet-101 81.69 44.65 DSSPN [58] VGG-16 76.04 34.56 DSSPN [58] ResNet-101 81.13 43.68 PSANet [34] ResNet-50 80.92 42.97 PSANet [34] ResNet-101 81.51 43.77 UperNet [59] ResNet-50 79.98 41.22 GANet [35] VGG-16 78.64 41.52 GANet [35] ResNet-50 81.80 44.71 GANet [35] ResNet-101 82.14 45.36 WResNet [53] ResNet-101 81.17 43.73 OCNet [32] ResNet-50 80.44 42.63 OCNet [32] ResNet-101 81.86 45.08 Fast-OCNet [32] ResNet jects, and the labels are inaccurate with ambiguity annotations. Thus, it mimicks a more natural object occurrence in daily scenes.…”

Section: Methodsmentioning

confidence: 99%

See 1 more Smart Citation

Deep Multiphase Level Set for Scene Parsing

Zhang

Liu

Lei

et al. 2020

IEEE Trans. on Image Process.

View full text Add to dashboard Cite

Recently, Fully Convolutional Network (FCN) seems to be the go-to architecture for image segmentation, including semantic scene parsing. However, it is difficult for a generic FCN to discriminate pixels around the object boundaries, thus FCN based methods may output parsing results with inaccurate boundaries. Meanwhile, level set based active contours are superior to the boundary estimation due to the sub-pixel accuracy that they achieve. However, they are quite sensitive to initial settings. To address these limitations, in this paper we propose a novel Deep Multiphase Level Set (DMLS) method for semantic scene parsing, which efficiently incorporates multiphase level sets into deep neural networks. The proposed method consists of three modules, i.e., recurrent FCNs, adaptive multiphase level set, and deeply supervised learning. More specifically, recurrent FCNs learn multi-level representations of input images with different contexts. Adaptive multiphase level set drives the discriminative contour for each semantic class, which makes use of the advantages of both global and local information. In each time-step of the recurrent FCNs, deeply supervised learning is incorporated for model training. Extensive experiments on three public benchmarks have shown that our proposed method achieves new state-of-the-art performances.

show abstract

Section: B Fcn Based Methodsmentioning

confidence: 99%

Section: Methodsmentioning

confidence: 99%

Deep Multiphase Level Set for Scene Parsing

Zhang

Liu

Lei

et al. 2020

IEEE Trans. on Image Process.

View full text Add to dashboard Cite

show abstract

“…MoE aims to learn a system composed of many separated networks (experts), where each expert learns to handle a subset of the whole dataset. Recently, deep MoE methods have shown their superiority in image recognition [1,20,51], machine translation [44], scene parsing [14] and so on. Unlike these works, we design a learnable voting network that can be updated with a novel meta-learning algorithm.…”

Section: Related Workmentioning

confidence: 99%

Generalizable Person Re-identification with Relevance-aware Mixture of Experts

Dai¹,

Li²,

Li³

et al. 2021

Preprint

View full text Add to dashboard Cite

Domain generalizable (DG) person re-identification (ReID) is a challenging problem because we cannot access any unseen target domain data during training. Almost all the existing DG ReID methods follow the same pipeline where they use a hybrid dataset from multiple source domains for training, and then directly apply the trained model to the unseen target domains for testing. These methods often neglect individual source domains' discriminative characteristics and their relevances w.r.t. the unseen target domains, though both of which can be leveraged to help the model's generalization. To handle the above two issues, we propose a novel method called the relevance-aware mixture of experts (RaMoE), using an effective voting-based mixture mechanism to dynamically leverage source domains' diverse characteristics to improve the model's generalization. Specifically, we propose a decorrelation loss to make the source domain networks (experts) keep the diversity and discriminability of individual domains' characteristics. Besides, we design a voting network to adaptively integrate all the experts' features into the more generalizable aggregated features with domain relevance. Considering the target domains' invisibility during training, we propose a novel learning-to-learn algorithm combined with our relation alignment loss to update the voting network. Extensive experiments demonstrate that our proposed RaMoE outperforms the state-of-the-art methods.

show abstract

“…Most deep learning based semantic segmentation methods are designed for RGB single-modality data, achieving prominent performance on many challenging large-scale datasets [11,12]. However, the RGB imaging sensors are highly sensitive to light and therefore susceptible to adverse lighting conditions, like darkness or overexposure.…”

Section: Introductionmentioning

confidence: 99%