Synthetic Infra-Red Image Evaluation Methods by Structural Similarity Index Measures

Article Computer Science, Artificial Intelligence

Loss functions for pose guided person image generation

Haoyue Shi et al.

Summary: This paper comprehensively studies the impact of different loss functions on pose guided person image generation, finding that a combination of adversarial loss, perceptual loss, and PSSIM loss yields optimal results.

PATTERN RECOGNITION (2022)

Add to Collection

Article Geochemistry & Geophysics

A Cascade Rotated Anchor-Aided Detector for Ship Detection in Remote Sensing Images

Ying Yu et al.

Summary: This article proposes a novel cascade rotated anchor-aided detection network for high-precision performance in detecting arbitrary-oriented ships in high-resolution remote sensing images. The network overcomes the limitations of state-of-the-art object detection methods applied directly to real ship detection through data preprocessing and a rotated anchor-aided detection module. Experimental results on challenging datasets demonstrate that the proposed method outperforms other methods.

IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING (2022)

Add to Collection

Review Computer Science, Information Systems

Conceptual and empirical comparison of dimensionality reduction algorithms (PCA, KPCA, LDA, MDS, SVD, LLE, ISOMAP, LE, ICA, t-SNE)

Farzana Anowar et al.

Summary: The study aims to explore feature extraction algorithms to address the curse of dimensionality in machine learning. It reviews various FEAs from different categories and evaluates their performance on challenging datasets.

COMPUTER SCIENCE REVIEW (2021)

Add to Collection

Article

Learning-based approach to underwater image dehazing using CycleGAN

Shima Ramesh Maniyath et al.

Arabian Journal of Geosciences (2021)

Add to Collection

Article Computer Science, Information Systems

Residual Learning of Cycle-GAN for Seismic Data Denoising

Wenda Li et al.

Summary: This paper proposes a cycle-GAN denoising framework based on data augmentation strategy to achieve intelligent seismic data denoising work, significantly reducing manual selection and intervention of denoising parameters. The method introduces residual learning into the cycle-GAN to enhance network training efficiency and uses a strategy to generate labeled datasets from unlabeled real noisy data. The proposed RCGAN method demonstrates excellent random noise suppression ability with minimal impact on useful seismic signals in both synthetic and real data experiments, showcasing effectiveness and superiority over state-of-the-art denoising methods.

IEEE ACCESS (2021)

Add to Collection

Article Biochemistry & Molecular Biology