Junping Yin

Speech-driven portrait animation generation models have made significant progress in generating realistic and dynamic portrait animations. The class of end-to-end latent diffusion paradigms represented by Hallo achieves impressive results in terms of alignment accuracy between audio inputs and visual outputs, encompassing lip movements, expressions and head poses. However, constrained by the suboptimal interaction design between reference portrait information and the denoising U-Net in such architectures, certain frames in the output video sequences suffer from inconsistencies in identity and background preservation. Moreover, the temporal attention within the temporal module operates by incorporating information across frames within each generation unit to capture overall motion trends, but ignoring shorter frame subsequences within the generation unit, consequently losing fine-grained details between adjacent frames. In order to solve the above problems, we take the end-to-end latent diffusion paradigm Hallo as the backbone, and construct a Multi-Source Self Attention (MSSA) to optimize the interaction between reference portrait identity information and denoising U-Net. In addition, we also propose a plug-and-play, training-free method known as Unit-wise Spectral-Blend Temporal Attention (U-SBTA), which enables simultaneously capture local high-frequency facial details from shorter frame subsequences within each generation unit, thereby improving facial fidelity in synthesized portrait videos. Our method is comprehensively evaluated on public dataset and our collected datasets from qualitative and quantitative analysis. The results demonstrate that the portrait animation videos generated by our method are better able to preserve identity and background consistency with the reference portrait, as well as exhibiting superior facial detail fidelity.

Accurate and efficient Video Polyp Segmentation (VPS) is vital for the early detection of colorectal cancer and the effective treatment of polyps. However, achieving this remains highly challenging due to the inherent difficulty in modeling the spatial-temporal relationships within colonoscopy videos. Existing methods that directly associate video frames frequently fail to account for variations in polyp or background motion, leading to excessive noise and reduced segmentation accuracy. Conversely, approaches that rely on optical flow models to estimate motion and align frames incur significant computational overhead. To address these limitations, we propose a novel VPS framework, termed Deformable Alignment and Local Attention (DALA). In this framework, we first construct a shared encoder to jointly encode the feature representations of paired video frames. Subsequently, we introduce a Multi-Scale Frame Alignment (MSFA) module based on deformable convolution to estimate the motion between reference and anchor frames. The multi-scale architecture is designed to accommodate the scale variations of polyps arising from differing viewing angles and speeds during colonoscopy. Furthermore, Local Attention (LA) is employed to selectively aggregate the aligned features, yielding more precise spatial-temporal feature representations. Extensive experiments conducted on the challenging SUN-SEG dataset and PolypGen dataset demonstrate that DALA achieves superior performance compared to stateof-the-art models. The code will be publicly available at https://github.com/xff12138/DALA.

This paper examines the asymmetric correlation and dynamic interaction between China’s macroeconomic market and the global crude oil market, addressing a critical limitation in existing literature: the frequency mismatch between high-frequency (daily) crude oil data and low-frequency (monthly) macroeconomic data. To resolve this, we employ a mixed-frequency group factor model that decomposes volatility drivers into two mutually exclusive components: (1) common factors, which capture cross-market spillovers between the two markets; and (2) group-specific factors, including low-frequency (LF)-specific factors (for China’s macroeconomic indicators) and high-frequency (HF)-specific factors (for crude oil prices). Our empirical analysis uses a comprehensive dataset spanning January 2005 to March 2024, covering 11 daily crude oil price indicators and 60 monthly Chinese macroeconomic indicators. We validate results using the adjusted coefficient of determination ( R ² ) and Bayesian Information Criterion (BIC) for model selection, and further test robustness across three samples: a full sample (2005.01–2024.03) and two crisis sub-samples (2007.01–2009.12 Financial Crisis, 2020.01–2023.12 COVID-19). Three core findings emerge: First, the two markets exhibit strong asymmetric influence; Second, the correlation is time-varying and crisis-sensitive; Third, factors show long-term persistence. These results confirm that crude oil acts as a key external constraint on China’s macroeconomic stability, while China’s macroeconomic conditions have limited impact on global oil pricing—consistent with its status as a “price taker” in the global crude oil market. The study provides empirical support for policymakers to design targeted risk-mitigation strategies and for market participants to optimize oil-related investment and risk management.

Clouds are ubiquitous in remote sensing images, and most of the existing methods for cloud removal are limited to either implementing multispectral images or exploiting supervised learning techniques. In this paper, we propose an unsupervised diffusion method by adapting the null space learning. The proposed method is built upon two trained denoising diffusion probabilistic models on diverse remote sensing datasets so as to cope with the mixed data from different sources. The matrices involving simplified degradation and "self-adaptive" generalized inverse are devised for the null space decomposition. For the diffusion model with null space decomposition, we derive its continuous reverse-time stochastic differential equation (SDE) and prove its variance-preserving property. Furthermore, we formulate explicitly the expectation of reverse-time SDE, which expedites the numerical efficiency of the proposed method. Numerical experiments on some remote sensing images are implemented to demonstrate the performance of the proposed method.