Gated Path Selection Network for Semantic Segmentation

Geng, Qichuan; Zhang, Hong; Qi, Xiaojuan; Huang, Gao; Yang, Ruigang; Zhou, Zhong

doi:10.1109/tip.2020.3046921

Cited by 29 publications

(16 citation statements)

References 35 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Recently, soft masks have been proposed [29] to adjust the contextual information, which can be seen as the combination of several normalized attention maps to dynamically select adaptive forward paths for every pixel. However, according to the mathematical analyses in terms of parameter optimization we shall provide in this paper, the direct functional relationships between the soft masks and corresponding features in existing pixel-wise dynamic structures [29] would make it difficult to obtain the global optimums.…”

Section: B Contextual Information Modelingmentioning

confidence: 99%

“…Although dynamic routing mentioned in Sec.III-A can adapt to scale distributions during the inference procedure, it rarely considers the diversity over different image areas, which is remarkable in remote sensing images due to the wide range of covered geographical areas. Recently proposed GPSNet [29] extends the idea of dynamic routing, where the utilized soft masks can be seen as the activation factor maps to select adaptive forward paths for each pixel. Specifically, we have…”

Section: B Gated Path Selectionmentioning

confidence: 99%

“…However, with all pixels sharing the same activating factors, such methods might ignore the diversity over different image areas, the effect of which is remarkable in remote sensing images due to the vast covered geographical areas with diverse land-cover distributions. Consequently, the Gated Path Selection Network (GPSNet) [29] has been proposed to select pixel-wise adaptive forward paths, simply shown at the top of Fig. 1, through designed soft masks instead of same activating factors shared by all pixels, as in [24].…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Hidden Path Selection Network for Semantic Segmentation of Remote Sensing Images

Yang,

Tong,

Xia

et al. 2021

Preprint

View full text Add to dashboard Cite

Targeting at depicting land covers with pixel-wise semantic categories, semantic segmentation in remote sensing images needs to portray diverse distributions over vast geographical locations, which is difficult to be achieved by the homogeneous pixel-wise forward paths in the architectures of existing deep models. Although several algorithms have been designed to select pixel-wise adaptive forward paths for natural image analysis, it still lacks theoretical supports on how to obtain optimal selections. In this paper, we provide mathematical analyses in terms of the parameter optimization, which guides us to design a method called Hidden Path Selection Network (HPS-Net). With the help of hidden variables derived from an extra mini-branch, HPS-Net is able to tackle the inherent problem about inaccessible global optimums by adjusting the direct relationships between feature maps and pixel-wise path selections in existing algorithms, which we call hidden path selection. For the better training and evaluation, we further refine and expand the 5-class Gaofen Image Dataset (GID-5) to a new one with 15 land-cover categories, i.e., GID-15. The experimental results on both GID-5 and GID-15 demonstrate that the proposed modules can stably improve the performance of different deep structures, which validates the proposed mathematical analyses.

show abstract

Section: B Contextual Information Modelingmentioning

confidence: 99%

Section: B Gated Path Selectionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Hidden Path Selection Network for Semantic Segmentation of Remote Sensing Images

Yang,

Tong,

Xia

et al. 2021

Preprint

View full text Add to dashboard Cite

show abstract

“…The advantages of deep learning present as high accuracy, good generalization performance, and strong anti-interference ability. It has obtained many successful industrial applications, such as face recognition [6], cityscape segmentation [7], style transfer [8] and defect detection [2], [9]. Deep learning algorithm will automatically extract the semantic and texture feature of the input image, without designing a specific feature extractor for different tasks.…”

Section: Introductionmentioning

confidence: 99%

SIGAN: A Novel Image Generation Method for Solar Cell Defect Segmentation and Augmentation

Su¹,

Zhou²,

Chen³

et al. 2021

Preprint

Self Cite

View full text Add to dashboard Cite

Solar cell electroluminescence (EL) defect segmentation is an interesting and challenging topic. Many methods have been proposed for EL defect detection, but these methods are still unsatisfactory due to the diversity of the defect and background. In this paper, we provide a new idea of using generative adversarial network (GAN) for defect segmentation. Firstly, the GAN-based method removes the defect region in the input defective image to get a defect-free image, while keeping the background almost unchanged. Then, the subtracted image is obtained by making difference between the defective input image with the generated defect-free image. Finally, the defect region can be segmented through thresholding the subtracted image. To keep the background unchanged before and after image generation, we propose a novel strong identity GAN (SIGAN), which adopts a novel strong identity loss to constraint the background consistency. The SIGAN can be used not only for defect segmentation, but also small-samples defective dataset augmentation. Moreover, we release a new solar cell EL image dataset named as EL-2019, which includes three types of images: crack, finger interruption and defect-free. Experiments on EL-2019 dataset show that the proposed method achieves 90.34% Fscore, which outperforms many state-of-the-art methods in terms of solar cell defects segmentation results.

show abstract

“…Semantic segmentation (Geng et al 2021;Yuan and Wang 2018) aims to assign a semantic class label for each pixel. These approaches are trained on a set of known semantic classes.…”

Section: Introductionmentioning

confidence: 99%

Anomaly Discovery in Semantic Segmentation via Distillation Comparison Networks

Zhou¹,

Gong²,

Zhou³

et al. 2021

Preprint

View full text Add to dashboard Cite

This paper aims to address the problem of anomaly discovery in semantic segmentation. Our key observation is that semantic classification plays a critical role in existing approaches, while the incorrectly classified pixels are easily regarded as anomalies. Such a phenomenon frequently appears and is rarely discussed, which significantly reduces the performance of anomaly discovery. To this end, we propose a novel Distillation Comparison Network (DiCNet). It comprises of a teacher branch which is a semantic segmentation network that removed the semantic classification head, and a student branch that is distilled from the teacher branch through a distribution distillation. We show that the distillation guarantees the semantic features of the two branches hold consistency in the known classes, while reflect inconsistency in the unknown class. Therefore, we leverage the semantic feature discrepancy between the two branches to discover the anomalies. DiCNet abandons the semantic classification head in the inference process, and hence significantly alleviates the issue caused by incorrect semantic classification. Extensive experimental results on StreetHazards dataset and BDD-Anomaly dataset are conducted to verify the superior performance of DiCNet. In particular, DiCNet obtains a 6.3% improvement in AUPR and a 5.2% improvement in FPR95 on StreetHazards dataset, achieves a 4.2% improvement in AUPR and a 6.8% improvement in FPR95 on BDD-Anomaly dataset. Codes are available at https://github.com/zhouhuan-hust/DiCNet.

show abstract

Gated Path Selection Network for Semantic Segmentation

Cited by 29 publications

References 35 publications

Hidden Path Selection Network for Semantic Segmentation of Remote Sensing Images

Hidden Path Selection Network for Semantic Segmentation of Remote Sensing Images

SIGAN: A Novel Image Generation Method for Solar Cell Defect Segmentation and Augmentation

Anomaly Discovery in Semantic Segmentation via Distillation Comparison Networks

Contact Info

Product

Resources

About