UWFormer: Underwater Image Enhancement via a Semi-Supervised Multi-Scale Transformer

2022 International Joint Conference on Neural Networks (IJCNN), Journal Year: 2024, Volume and Issue: 48, P. 1 - 8

Published: June 30, 2024

Language: Английский

---

3D Wasserstein Generative Adversarial Network with Dense U-Net-Based Discriminator for Preclinical fMRI Denoising

Deleted Journal, Journal Year: 2025, Volume and Issue: unknown

Published: Feb. 12, 2025

Functional magnetic resonance imaging (fMRI) is extensively used in clinical and preclinical settings to study brain function; however, fMRI data inherently noisy due physiological processes, hardware, external noise. Denoising one of the main preprocessing steps any analysis pipeline. This process challenging comparison variations geometry, image resolution, low signal-to-noise ratios. In this paper, we propose a structure-preserved algorithm based on 3D Wasserstein generative adversarial network with dense U-net-based discriminator called U-WGAN. We apply 4D configuration effectively denoise temporal spatial information analyzing data. GAN-based denoising methods often utilize identify significant differences between denoised noise-free images, focusing global or local features. To refine model, our method employs U-Net learn both distinctions. tackle potential oversmoothing, introduce an loss enhance perceptual similarity by measuring feature space distances. Experiments illustrate that U-WGAN significantly improves quality resting-state task data, enhancing ratio without introducing excessive structural changes existing methods. The proposed outperforms state-of-the-art when applied simulated real

Language: Английский

---

ShaDocFormer: A Shadow-Attentive Threshold Detector With Cascaded Fusion Refiner for Document Shadow Removal

2022 International Joint Conference on Neural Networks (IJCNN), Journal Year: 2024, Volume and Issue: unknown, P. 1 - 8

Published: June 30, 2024

Language: Английский

---

Aeroengine Bearing Time-Varying Skidding Assessment With Prior Knowledge-Embedded Dual Feedback Spatial-Temporal GCN

IEEE Transactions on Cybernetics, Journal Year: 2024, Volume and Issue: 55(2), P. 826 - 839

Published: Nov. 20, 2024

Bearing skidding is the primary factor restricting development of aeroengines toward ultrahigh speed, low friction, and lightweight. Compared to typical bearing faults, analysis presents greater challenges due weak signal properties, significant time-varying characteristics coupling influence multiple factors. It crucial fully utilize multisource signals enhance features capture characteristics. This article proposes a prior knowledge-embedded dual feedback spatial-temporal graph convolutional network (DFSTGCN) for assessment. Unlike existing adjacency matrix construction strategies, correlation between described based on knowledge, which includes dynamic model, structural dynamics, expert experience. Furthermore, DFSTGCN designed simultaneously focus spatial temporal dependencies data. Specifically, mechanism that prediction error ratio uncertainty loss function employed improve generalization performance model. The effectiveness proposed strategy validated under different working conditions.

Language: Английский

---

Redefining QUAIDE: paving the way for better AI in preclinical endoscopy

Gut, Journal Year: 2024, Volume and Issue: unknown, P. gutjnl - 334324

Published: Dec. 20, 2024

Language: Английский

---