Detail-recovery Image Deraining via Context Aggregation Networks

Deng, Sen; Wei, Mingqiang; Wang, Jun; Feng, Yidan; Liang, Luming; Xie, Haoran; Wang, Fu Lee; Wang, Meng

doi:10.1109/cvpr42600.2020.01457

“…MPRNet [50] CCN [34] Semi-DerainGAN [58] CVID [59] RCDNet [53] Pan et al [60] MSPFN [44] RDDAN [61] JDNet [45] QuDeC [62] Syn2Real [ [70] GraNet [46] LPNet [71] PReNet [31] DAF-Net [22] ReHEN [72] DDC-Net [35] RR-GAN [73] MH-DerainNet [74] SIRR [75] SPANet [33] ReMAEN [76] UD-GAN [77] UMRL [78] ID-CGAN [32] Li et al [23] JORDER-E [56] JORDER [21] RWL [79] DualCNN [80] NLEDN [81] RESCAN [55] ResGuideNet [82] Qian et al [25] DID-MDN [24] Li et al [83] DerainNet [83] Fu et al [20] Quan et al [52] SSDRNet [67] MOEDN [56] RICNet [35] JRGR [57] RLNet [59] QSMD [58] Fig. 7: The division of recent SID methods from six aspects based on the three factors.…”

Section: Synthetical Mathematical General Specificmentioning

confidence: 99%

“…In addition, by utilizing unique modules, SID methods can obtain more representational features. Some of these modules are universal modules, such as dilated convolution [98] used in [19,55], LSTM [94] used in [25,26], U-net [96] used in [26,41], SENet [99] used in [24,72], and ShuffleNet [100] used in QSMD [36]. Others are based on attention [102] mechanism and specifically designed, such as the spatial attentive module used in SPANet [27] and confidence map network used in UMRL [83].…”

Section: Domain Knowledgementioning

confidence: 99%

Data-Driven Single Image Deraining: A Comprehensive Review and New Perspectives

Zhang¹,

Yue²,

Zhang³

et al. 2023

Preprint

Self Cite

1

0

View full text Add to dashboard Cite

<div>Abstract—Due to the powerful fitting ability of neural networks and massive training data, data-driven Single Image Deraining (SID) methods have obtained significant performance, and most of the existing studies focus on improving the deraining performance by proposing different kinds of deraining networks. However, the generalization ability of current SID methods may still be limited in the real scenario, and the deraining results also cannot effectively improve the performance of subsequent high-level tasks (such as object detection). The main reason may be because the current works mainly focus on designing new deraining neural networks, while neglecting the interpretation of the solving process. To investigate these issues, we in this paper re-examine the three important factors (i.e.,data, rain model and network architecture) for the SID task, and specifically analyze them by proposing new and more reasonable criteria (i.e., general vs. specific, synthetical vs. mathematical, black-box vs. white-box). We will also study the relationship of the three factors from new perspectives of data, and reveal two different solving paradigms (explicit vs. implicit) for the SID task. Besides, we also profoundly studied one of the three factors, i.e., data, and measured the performance of current methods on different datasets via extensive experiments to reveal the effectiveness of SID data. Finally, with the above comprehensive review and in-depth analysis, we also draw some valuable conclusions and suggestions for future research.</div>

show abstract

“…MPRNet [50] CCN [34] Semi-DerainGAN [58] CVID [59] RCDNet [53] Pan et al [60] MSPFN [44] RDDAN [61] JDNet [45] QuDeC [62] Syn2Real [ [70] GraNet [46] LPNet [71] PReNet [31] DAF-Net [22] ReHEN [72] DDC-Net [35] RR-GAN [73] MH-DerainNet [74] SIRR [75] SPANet [33] ReMAEN [76] UD-GAN [77] UMRL [78] ID-CGAN [32] Li et al [23] JORDER-E [56] JORDER [21] RWL [79] DualCNN [80] NLEDN [81] RESCAN [55] ResGuideNet [82] Qian et al [25] DID-MDN [24] Li et al [83] DerainNet [83] Fu et al [20] Quan et al [52] SSDRNet [67] MOEDN [56] RICNet [35] JRGR [57] RLNet [59] QSMD [58] Fig. 7: The division of recent SID methods from six aspects based on the three factors.…”

Section: Synthetical Mathematical General Specificmentioning

confidence: 99%

“…In addition, by utilizing unique modules, SID methods can obtain more representational features. Some of these modules are universal modules, such as dilated convolution [98] used in [19,55], LSTM [94] used in [25,26], U-net [96] used in [26,41], SENet [99] used in [24,72], and ShuffleNet [100] used in QSMD [36]. Others are based on attention [102] mechanism and specifically designed, such as the spatial attentive module used in SPANet [27] and confidence map network used in UMRL [83].…”

Section: Domain Knowledgementioning

confidence: 99%

Data-Driven Single Image Deraining: A Comprehensive Review and New Perspectives

Zhang¹,

Yue²,

Zhang³

et al. 2021

Preprint

Self Cite

View full text Add to dashboard Cite

<div>Abstract—Due to the powerful fitting ability of neural networks and massive training data, data-driven Single Image Deraining (SID) methods have obtained significant performance, and most of the existing studies focus on improving the deraining performance by proposing different kinds of deraining networks. However, the generalization ability of current SID methods may still be limited in the real scenario, and the deraining results also cannot effectively improve the performance of subsequent high-level tasks (such as object detection). The main reason may be because the current works mainly focus on designing new deraining neural networks, while neglecting the interpretation of the solving process. To investigate these issues, we in this paper re-examine the three important factors (i.e.,data, rain model and network architecture) for the SID task, and specifically analyze them by proposing new and more reasonable criteria (i.e., general vs. specific, synthetical vs. mathematical, black-box vs. white-box). We will also study the relationship of the three factors from new perspectives of data, and reveal two different solving paradigms (explicit vs. implicit) for the SID task. Besides, we also profoundly studied one of the three factors, i.e., data, and measured the performance of current methods on different datasets via extensive experiments to reveal the effectiveness of SID data. Finally, with the above comprehensive review and in-depth analysis, we also draw some valuable conclusions and suggestions for future research.</div>

show abstract

“…In addition, by utilizing unique modules, SID methods can obtain more representational features. Some of these modules are universal modules, such as dilated convolution [99] used in [19,55], LSTM [95] used in [25,26], U-net [97] used in [26,41], SENet [100] used in [24,73], and ShuffleNet [101] used in QSMD [38]. Others are based on attention [102] mechanism and specifically designed, such as the spatial attentive module used in SPANet [27] and confidence map network used in UMRL [84].…”

Section: Domain Knowledgementioning

confidence: 99%

Data-Driven Single Image Deraining: A Comprehensive Review and New Perspectives

Zhang¹,

Yue²,

Zhang³

et al. 2021

Preprint

Self Cite

1

0

View full text Add to dashboard Cite

<div>Abstract—Due to the powerful fitting ability of neural networks and massive training data, data-driven Single Image Deraining (SID) methods have obtained significant performance, and most of the existing studies focus on improving the deraining performance by proposing different kinds of deraining networks. However, the generalization ability of current SID methods may still be limited in the real scenario, and the deraining results also cannot effectively improve the performance of subsequent high-level tasks (such as object detection). The main reason may be because the current works mainly focus on designing new deraining neural networks, while neglecting the interpretation of the solving process. To investigate these issues, we in this paper re-examine the three important factors (i.e.,data, rain model and network architecture) for the SID task, and specifically analyze them by proposing new and more reasonable criteria (i.e., general vs. specific, synthetical vs. mathematical, black-box vs. white-box). We will also study the relationship of the three factors from new perspectives of data, and reveal two different solving paradigms (explicit vs. implicit) for the SID task. Besides, we also profoundly studied one of the three factors, i.e., data, and measured the performance of current methods on different datasets via extensive experiments to reveal the effectiveness of SID data. Finally, with the above comprehensive review and in-depth analysis, we also draw some valuable conclusions and suggestions for future research.</div>

show abstract

Detail-recovery Image Deraining via Context Aggregation Networks

Cited by 181 publications

References 26 publications

Data-Driven Single Image Deraining: A Comprehensive Review and New Perspectives

Data-Driven Single Image Deraining: A Comprehensive Review and New Perspectives

Data-Driven Single Image Deraining: A Comprehensive Review and New Perspectives

Data-Driven Single Image Deraining: A Comprehensive Review and New Perspectives

Contact Info

Product

Resources

About