Effective Use of Dilated Convolutions for Segmenting Small Object Instances in Remote Sensing Imagery

Hamaguchi, Ryuhei; Fujita, Aito; Nemoto, Keisuke; Imaizumi, Tsutomu; Hikosaka, Shuhei

doi:10.1109/wacv.2018.00162

Cited by 213 publications

(167 citation statements)

References 13 publications

Supporting

Mentioning

165

Contrasting

Unclassified

Order By: Relevance

“…2. This structure is similar to the basic context network architecture proposed in [14], but we add more dilated convolution layers with decreasing l. This improves local low-level features, where spatial relationships amongst adjacent pixels may be ignored due to sparsity of the dilated filters when increasing l [15].…”

Section: Dilated Convolutionmentioning

confidence: 99%

Defectnet: Multi-Class Fault Detection on Highly-Imbalanced Datasets

Anantrasirichai

Bull

2019

2019 IEEE International Conference on Image Processing (ICIP)

View full text Add to dashboard Cite

As a data-driven method, the performance of deep convolutional neural networks (CNN) relies heavily on training data. The prediction results of traditional networks give a bias toward larger classes, which tend to be the background in the semantic segmentation task. This becomes a major problem for fault detection, where the targets appear very small on the images and vary in both types and sizes. In this paper we propose a new network architecture, DefectNet, that offers multiclass (including but not limited to) defect detection on highlyimbalanced datasets. DefectNet consists of two parallel paths, which are a fully convolutional network and a dilated convolutional network to detect large and small objects respectively. We propose a hybrid loss maximising the usefulness of a dice loss and a cross entropy loss, and we also employ the leaky rectified linear unit (ReLU) to deal with rare occurrence of some targets in training batches. The prediction results show that our DefectNet outperforms state-of-the-art networks for detecting multi-class defects with the average accuracy improvement of approximately 10% on a wind turbine.

show abstract

Section: Dilated Convolutionmentioning

confidence: 99%

Defectnet: Multi-Class Fault Detection on Highly-Imbalanced Datasets

Anantrasirichai

Bull

2019

2019 IEEE International Conference on Image Processing (ICIP)

View full text Add to dashboard Cite

show abstract

“…Dilated convolutions help the network model to predict instances without losing receptive field, without needs of the fully convolutional layer, and with less learnable parameters in comparison to previous FCN methods. Existence of objects at multiple scales is another challenge in semantic segmentation (L.-C. C. which many recent studies have been focused on solving that issue (Hamaguchi et al, 2018;Lin et al, 2017). Few studies have been applied deep semantic segmentation for the problem of ultrasound tongue extraction.…”

Section: Literature Review and Related Workmentioning

confidence: 99%

“…Using dilated convolution (sometimes called atrous convolution) while the dilation factor is increased monotonously through layers, it revealed that the receptive field could be effectively expanded with keeping a spatial resolution. However, the sparsity of dilated kernels does not always cause performance improvement especially for small objects and details (Hamaguchi et al, 2018). To solve this problem, one solution is to decrease the dilation factor throughout the decoding path of the network model similar to the number of kernels in the UNet network (Hamaguchi et al, 2018).…”

Section: Dilated Convolutionmentioning

confidence: 99%

“…However, the sparsity of dilated kernels does not always cause performance improvement especially for small objects and details (Hamaguchi et al, 2018). To solve this problem, one solution is to decrease the dilation factor throughout the decoding path of the network model similar to the number of kernels in the UNet network (Hamaguchi et al, 2018). Figure 2 illustrates samples of decoding strategies using interpolation, deconvolution, and dilated convolution.…”

Section: Dilated Convolutionmentioning

confidence: 99%

“…Drop-out layers and batch normalization layers are followed convolutional layers to improve regularization and accuracy of network models. From Table 1, dilation factor is first increased from one to eight and then decreased again to one in a forward path of successive dilated convolutions to solve the problem of spatial inconsistency (Hamaguchi et al, 2018). It is common in deep learning semantic segmentation to use SoftMax for the last layer of the network, and at least part of the system is a modified version of VGG16 network (Badrinarayanan et al, 2015;Long et al, 2015;Simonyan and Zisserman, 2014).…”

Section: Network Architecturesmentioning

confidence: 99%

See 2 more Smart Citations

BowNet: Dilated convolutional neural network for ultrasound tongue contour extraction

Mozaffari

Lee²

2019

The Journal of the Acoustical Society of America

View full text Add to dashboard Cite

Ultrasound imaging is safe, relatively affordable, and capable of real-time performance. This technology has been used for real-time visualization and analyzing the functionality of human organs in many studies. One application of this technology is to visualize and to characterize human tongue shape and motion during a real-time speech to study healthy or impaired speech production. Due to the noisy nature of ultrasound images with low-contrast characteristic, it might require expertise for non-expert users to recognize organ shape such as tongue surface (dorsum). To alleviate this difficulty for quantitative analysis of tongue shape and motion, tongue surface can be extracted, tracked, and visualized instead of the whole tongue region. Delineating the tongue surface from each frame is a cumbersome, subjective, and error-prone task. Furthermore, the rapidity and complexity of tongue gestures have made it a challenging task, and manual segmentation is not a feasible solution for real-time applications. The progress of deep convolutional neural networks has been successfully exploited in various computer vision tasks such as image classification and segmentation. Several end-to-end deep learning segmentation methods provide a promising alternative for previous techniques with higher accuracy and robustness results, without any intervention. Employing the power of highspeed graphics processing unit (GPU) with state-of-the-art deep neural network models and training techniques, it is feasible to implement new fully-automatic, accurate, and robust segmentation methods with the capability of real-time performance, applicable for tracking of the tongue contours during the speech. This paper presents two novel deep neural network models named BowNet and wBowNet benefits from the ability of global prediction of decoding-encoding models, with integrated multi-scale contextual information, and capability of full-resolution (local) extraction of dilated convolutions. Experimental results using several ultrasound tongue image datasets revealed that the combination of both localization and globalization searching could improve prediction result significantly. Assessment of BowNet models using both qualitatively and quantitatively studies showed their outstanding achievements in terms of accuracy and robustness in compare with similar techniques.

show abstract

FFNet: A simple image dedusting network with feature fusion

Huang

Wang

et al. 2021

Concurrency and Computation

View full text Add to dashboard Cite

Dust is a common air pollution source. The color of images captured under dusty weather is usually yellow even brown, which reduces scene visibility and causes the loss of image details. To remove the dust and make image scene clear, this article presents a simple and effective image dedusting network called FFNet. In the process of image feature extraction, the FFNet uses several residual blocks with smoothed dilated convolution, that is, common dilated convolution followed by separable and shared (SS) blockwise fully connected operation to extend the receptive field and reduce gridding artifacts caused by common dilated convolution. Furthermore, the FFNet fuses image features from different layers via an adaptive weighting scheme. Due to the difficulty of collecting real dusty images, we used our proposed dusty image synthesis scheme to achieve data augmentation for better network training. Experiments on a series of synthetic and real dusty images demonstrate that the FFNet obtains better image dedusting performance than several state-of-the-art image restoration methods.

show abstract

Effective Use of Dilated Convolutions for Segmenting Small Object Instances in Remote Sensing Imagery

Cited by 213 publications

References 13 publications

Defectnet: Multi-Class Fault Detection on Highly-Imbalanced Datasets

Defectnet: Multi-Class Fault Detection on Highly-Imbalanced Datasets

BowNet: Dilated convolutional neural network for ultrasound tongue contour extraction

FFNet: A simple image dedusting network with feature fusion

Contact Info

Product

Resources

About