arXiv.itsnick.co

an independent field guide to 23 million papers from arXiv & OpenAlex — not affiliated with arXiv.org

Computer Science · Computer Vision and Pattern Recognition

De-raining using Dual Graph Convolutional Network

K. Ashwini, Gade Rithish Reddy, Kallem Bhargava Srinivas

· 2022 IEEE International Conference on Distributed Computing and Electrical Circuits and Electronics (ICDCECE)

See on the map

Abstract

Drastic climatic conditions, such as rain, snow, and other extreme weather, have a negative impact on the visual quality of photographs taken, rendering them unusable for use and sharing. As a result, it’s critical to develop a solution to this type of single- image de-raining problem. So, we try to apply deep CNN-based de-raining algorithms, which are built by stacking vanilla convolutional layers that can only simulate local relations. As a result, for this specific activity, long-range contextual information is rarely considered. To overcome these issues when training the system, we’d want to use a neural network to retrieve rain stripes from a single picture. called "Dual Graph Convolutional Network. "We incorporate these two graphs, as well as multi-scale dilated convolution, into symmetrically skip- connected network architecture to perform image rain removal. As a result, by exploring numerous representations, such as multi- scale local feature, global spatial coherence, and cross-channel correlation, this dual graph convolutional network can handle complicated and geographically extended rain streaks. Meanwhile, this model is simple to implement, trainable from beginning to end, and computationally efficient. Extensive tests on simulated and actual data show that our method outperforms current state-of-the- art methods by a wide margin.

More like this