Multi-Scale Neural Network With Dilated Convolutions for Image Deblurring

Ople, Jose Jaena Mari; Yeh, Pin-Yi; Sun, Shih-Wei; Tsai, I. F.; Hua, Kai-Lung

doi:10.1109/access.2020.2980996

Cited by 18 publications

(6 citation statements)

References 25 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Existing methods usually use standard convolution with fixed kernel size throughout the network module to model the temporal information (Yan et al, 2018 ; Li M. et al, 2019 ; Shi et al, 2019a , b ). In this paper, we proposed a multiscale aggregation learning method by introducing hybrid dilation convolution to improve the traditional temporal convolution module (Ople et al, 2020 ). Because of the exponential expansion of the perceptual field with guaranteed coverage, the proposed MS-TGCN can effectively aggregate multi-scale contextual information without loss of resolution by using dilation convolution.…”

Section: Proposed Methodsmentioning

confidence: 99%

Multi-scale and attention enhanced graph convolution network for skeleton-based violence action recognition

et al. 2022

View full text Add to dashboard Cite

Graph convolution networks (GCNs) have been widely used in the field of skeleton-based human action recognition. However, it is still difficult to improve recognition performance and reduce parameter complexity. In this paper, a novel multi-scale attention spatiotemporal GCN (MSA-STGCN) is proposed for human violence action recognition by learning spatiotemporal features from four different skeleton modality variants. Firstly, the original joint data are preprocessed to obtain joint position, bone vector, joint motion and bone motion datas as inputs of recognition framework. Then, a spatial multi-scale graph convolution network based on the attention mechanism is constructed to obtain the spatial features from joint nodes, while a temporal graph convolution network in the form of hybrid dilation convolution is designed to enlarge the receptive field of the feature map and capture multi-scale context information. Finally, the specific relationship in the different skeleton data is explored by fusing the information of multi-stream related to human joints and bones. To evaluate the performance of the proposed MSA-STGCN, a skeleton violence action dataset: Filtered NTU RGB+D was constructed based on NTU RGB+D120. We conducted experiments on constructed Filtered NTU RGB+D and Kinetics Skeleton 400 datasets to verify the performance of the proposed recognition framework. The proposed method achieves an accuracy of 95.3% on the Filtered NTU RGB+D with the parameters 1.21M, and an accuracy of 36.2% (Top-1) and 58.5% (Top-5) on the Kinetics Skeleton 400, respectively. The experimental results on these two skeleton datasets show that the proposed recognition framework can effectively recognize violence actions without adding parameters.

show abstract

Section: Proposed Methodsmentioning

confidence: 99%

Multi-scale and attention enhanced graph convolution network for skeleton-based violence action recognition

et al. 2022

View full text Add to dashboard Cite

show abstract

“…At the same time, we combine the mutual dual-channel energy loss function proposed in Chap. 3 to propose the nal deblurring model DeblurGAN + in this paper [14].…”

Section: Related Workmentioning

confidence: 99%

Application of image deblurring method and voice network analysis in 3D environment art design system

Xia

2023

Preprint

View full text Add to dashboard Cite

In a non-uniform blurred scene, pixels in other places in the blurred image undergo a downgrading process, which makes it difficult to accurately estimate the blur kernel. The method based on deep learning can directly realize image deblurring without estimating the blur kernel. Therefore, in this article, we will use deep learning methods to study the problem of blind listeners in dynamic scenes. Looking at the recent deep learning methods applied to denoisers, most image deblurring processing is set to image mapping. As the many advantages of network voice communication have quickly become an important part of people's online life, more and more researchers have begun to use voice network analysis as a steganography carrier. Compared with the old steganographic carrier, the use of network voice stream as the carrier has the advantages of immediacy, higher steganographic bandwidth and variable carrier length. However, steganography based on voice network analysis is combined with certain secure communication technologies to send harmful confidential information. This may become a major security threat because it is difficult to detect. Using 3D environment visualization technology, people can directly manipulate the physical information in the 3D graphics world and communicate directly with the computer. This 3D environment visualization technology integrates the power of humans and machines in an intuitive and natural way. These innovative changes will undoubtedly significantly improve people's work efficiency. Visualization technology enables people to interact with the art design system in real time, so people can use the art design system to obtain information or use previously unimaginable ways, and then they can play their creative thinking.

show abstract

“…Increasing the receptive field is also necessary in other areas of deep learning, and many works have been carried out in this regard. For instance, dilated convolution has been employed to deblur images [26], denoise images [27], classify images [28] and estimate the density of people [29]. To extract global features, it is crucial to use convolution kernels with different scales, as using a single 3 × 3 convolution kernel is not effective in obtaining global features.…”

Section: Related Workmentioning

confidence: 99%

MFF: An effective method of solving the ill regions in stereo matching

Ren

Feng

Xie

et al. 2023

IET Computer Vision

View full text Add to dashboard Cite

In the current stereo matching field, the accuracy of the derived disparity map is highly dependent on the processing capability of ill regions. Fortunately, we find that the use of local information will eliminate the negative effects associated with ill regions. As a result of the above discovery, we propose the Concatenated Dilated Convolution (CDC) block and the Multi‐scale Feature Fusion module (MFF), which are capable of effectively extracting regional context by increasing the receptive field in parallel and channel‐wise ways. The CDC block can expand the receptive field by applying multiple dilated convolutions at different dilation rates in parallel to enhance the smoothness of the feature map. By constructing parallel CDC blocks in a multiple dilated manner, the MFF module can improve the smoothness of the feature map. In addition, to control the number of parameters in the MFF network, a high‐performance channel‐distribution algorithm is proposed, capable of adjusting the weights of each module and convolution in an adaptive manner while reducing the number of parameters. Extensive experiments have demonstrated that MFF and CDC can effectively improve the performance of ill areas and networks with a minimal number of parameters.

show abstract

Multi-Scale Neural Network With Dilated Convolutions for Image Deblurring

Cited by 18 publications

References 25 publications

Multi-scale and attention enhanced graph convolution network for skeleton-based violence action recognition

Multi-scale and attention enhanced graph convolution network for skeleton-based violence action recognition

Application of image deblurring method and voice network analysis in 3D environment art design system

MFF: An effective method of solving the ill regions in stereo matching

Contact Info

Product

Resources

About