Xianhui Huang

Mr. Xianhui Huang

orcid

Research Focus & Metrics

Academic Career Length

6Years (2021-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
16 February 2025 in Electronics
Verified
Journal Article
Open access
PDF file
GLGuangping Li
Guangping Li
YGYanan Gao
Yanan Gao
XHXianhui Huang
Xianhui Huang
Wing-Kuen LingWing-Kuen Ling
Wing-Kuen Ling
Contrastive learning has emerged as a dominant paradigm for understanding 3D open-world environments, particularly in the realm of multi-modalities. However, due to the nature of self-supervised learning and the limited size of 3D datasets, pre-trained models in the 3D point cloud domain often suffer from overfitting in downstream tasks, especially in zero-shot classification. To tackle this problem, we design a module to mine and enhance hard negatives from datasets, which are useful to improve the discrimination of models. This module could be seamlessly integrated into cross-modal contrastive learning frameworks, addressing the overfitting issue by enhancing the mined hard negatives during the process of training. This module consists of two key components: mining and enhancing. In the process of mining, we identify hard negative samples by examining similarity relationships between vision–vision and vision–text modalities, locating hard negative pairs within the visual domain. In the process of enhancing, we compute weighting coefficients via the similarity differences of these mined hard negatives. By enhancing the mined hard negatives while leaving others unchanged, we improve the overall performance and discrimination of models. A series of experiments demonstrate that our module can be easily incorporated into various contrastive learning frameworks, leading to improved model performance in both zero-shot and few-shot tasks.

24 February 2026 in IEEE Sensors Journal
Verified
Journal Article
Closed access
GLGuangping Li
Guangping Li
XHXianhui Huang
Xianhui Huang
YGYanan Gao
Yanan Gao
Attention-based multi-modal fusion frameworks have significantly advanced View-guided Point Cloud Completion (ViPC), which aims to recover a complete 3D point cloud from a partial one with the aid of a single-view image. However, existing methods commonly employ asymmetric cross-attention for point cloud and image feature fusion, in which one modality is designated as the query and the other as key-value pairs, resulting in an information flow biased toward the querying modality rather than enabling flexible feature extraction across modalities. Moreover, supplement completion loss that require the participation of all point cloud and image tokens introduce substantial computational overhead. To address these limitations, we propose Memory Transfer with Contrastive Learning Network termed MTCLnet, a novel framework that leverages contrastive learning to effectively align the feature spaces between point cloud and image modalities. Furthermore, we design a cross-attention-based module equipped with learnable tokens to adaptively fuse informative features across modalities. The integrated design enables more accurate and computationally efficient point cloud completion. Extensive experiments on the ShapeNet-ViPC dataset demonstrate that our method outperforms existing approaches, achieving both improved reconstruction quality and reduced computational complexity.

22 April 2021 in Frontiers in Earth Science
Verified
Journal Article
Open access
PDF file
XHXianhui Huang
Xianhui Huang
YXYungui Xu
Yungui Xu
HLHaoyuan Li
Haoyuan Li
ZZZhiwen Zhang
Zhiwen Zhang
WXWei Xu
Wei Xu
The Moxi area in the Sichuan Basin is dominated by carbonate gas reservoirs, where gas productivity is most strongly influenced by their pore types. Fractured caves are the most favorable pore structure type for reservoir productivity, followed by cave and vuggy pore structures, and interparticle pore structures are the least productive. The spatial discrimination of these three pore types is important for cost-effective development. However, the pore type identification remains difficult owing to poor-quality azimuthal seismic data. A practical approach is to understand the seismic signatures of the different pore types and the related productivities from the post-stack data. In this work, seismic forward modeling is conducted using a constructed theoretical model of Hudson's anisotropic representation, and the pre-stack and post-stack anisotropy signatures are analyzed for different pore types. The rock model is further calibrated using log data, and forward modeling is performed based on the calibrated logs. We propose a new attribute of these signatures: namely, the ratio of the absolute peak and the absolute trough immediately below the peak, which is applied to the three-dimensional seismic data in the Moxi area. In contrast with other conventional attributes, this ratio effectively correlates with pore type, which allows the pore types in wells to be differentiated. This attribute also reasonably correlates with open flow gas rate of the well. The results demonstrate that this attribute from the post-stack data is a promising indicator of pore type and gas productivity and can also be readily mapped spatially for the selection of new drilling locations.