检索结果-内蒙古大学图书馆

Modality-aware triplet hard mining for zero-shot sketch-based image retrieval

学校读者我要写书评

暂无评论

arXiv 2021年

作者： Huang, Zongheng Sun, Yifan Han, Chuchu Gao, Changxin Sang, Nong Key Laboratory of Ministry of Education for Image Processing and Intelligent Control School of Artificial Intelligence and Automation Huazhong University of Science and Technology China Baidu Research United States

This paper tackles the Zero-Shot Sketch-Based image Retrieval (ZS-SBIR) problem from the viewpoint of cross-modality metric learning. This task has two characteristics: 1) the zero-shot setting requires a metric space with good within-class compactness and the between-class discrepancy for recognizing the novel classes and 2) the sketch query and the photo gallery are in different modalities. The metric learning viewpoint benefits ZS-SBIR from two aspects. First, it facilitates improvement through recent good practices in deep metric learning (DML). By combining two fundamental learning approaches in DML, e.g., classification training and pairwise training, we set up a strong baseline for ZS-SBIR. Without bells and whistles, this baseline achieves competitive retrieval accuracy. Second, it provides an insight that properly suppressing the modality gap is critical. To this end, we design a novel method named Modality-Aware Triplet Hard Mining (MATHM). MATHM enhances the baseline with three types of pairwise learning, e.g., a cross-modality sample pair, a within-modality sample pair, and their combination. We also design an adaptive weighting method to balance these three components during training dynamically. Experimental results confirm that MATHM brings another round of significant improvement based on the strong baseline and sets up new state-of-the-art performance. For example, on the TU-Berlin dataset, we achieve 47.88+2.94% mAP@all and 58.28+2.34% Prec@100. Code will be publicly available. Copyright © 2021, The Authors. All rights reserved.

关键词： Deep learning

Dual-Camera All-in-Focus Neural Radiance Fields

学校读者我要写书评

暂无评论

arXiv 2025年

作者： Luo, Xianrui Wu, Zijin Peng, Juewen Sun, Huiqiang Cao, Zhiguo Lin, Guosheng Key Laboratory of Image Processing and Intelligent Control Ministry of Education School of Artificial Intelligence and Automation Huazhong University of Science and Technology Wuhan430074 China Singapore639798 Singapore

We present the first framework capable of synthesizing the all-in-focus neural radiance field (NeRF) from inputs without manual refocusing. Without refocusing, the camera will automatically focus on the fixed object for all views, and current NeRF methods typically using one camera fail due to the consistent defocus blur and a lack of sharp reference. To restore the all-in-focus NeRF, we introduce the dual-camera from smartphones, where the ultra-wide camera has a wider depth-of-field (DoF) and the main camera possesses a higher resolution. The dual camera pair saves the high-fidelity details from the main camera and uses the ultra-wide camera’s deep DoF as reference for all-in-focus restoration. To this end, we first implement spatial warping and color matching to align the dual camera, followed by a defocus-aware fusion module with learnable defocus parameters to predict a defocus map and fuse the aligned camera pair. We also build a multi-view dataset that includes image pairs of the main and ultra-wide cameras in a smartphone. Extensive experiments on this dataset verify that our solution, termed DC-NeRF, can produce high-quality all-in-focus novel views and compares favorably against strong baselines quantitatively and qualitatively. We further show DoF applications of DC-NeRF with adjustable blur intensity and focal plane, including refocusing and split diopter. Copyright © 2025, The Authors. All rights reserved.

关键词：

UniAnimate-DiT: Human image Animation with Large-Scale Video Diffusion Transformer

学校读者我要写书评

暂无评论

arXiv 2025年

作者： Wang, Xiang Zhang, Shiwei Tang, Longxiang Zhang, Yingya Gao, Changxin Wang, Yuehuan Sang, Nong Key Laboratory of Image Processing and Intelligent Control School of Artificial Intelligence and Automation Huazhong University of Science and Technology United States Alibaba Group China Tsinghua University China

This report presents UniAnimate-DiT, an advanced project that leverages the cutting-edge and powerful capabilities of the open-source Wan2.1 model for consistent human image animation. Specifically, to preserve the robust generative capabilities of the original Wan2.1 model, we implement Low-Rank Adaptation (LoRA) technique to fine-tune a minimal set of parameters, significantly reducing training memory overhead. A lightweight pose encoder consisting of multiple stacked 3D convolutional layers is designed to encode motion information of driving poses. Furthermore, we adopt a simple concatenation operation to integrate the reference appearance into the model and incorporate the pose information of the reference image for enhanced pose alignment. Experimental results show that our approach achieves visually appearing and temporally consistent high-fidelity animations. Trained on 480p (832x480) videos, UniAnimate-DiT demonstrates strong generalization capabilities to seamlessly upscale to 720P (1280x720) during inference. The training and inference code is publicly available at https://github. com/ali-vilab/UniAnimate-DiT. Copyright © 2025, The Authors. All rights reserved.

关键词：

Prognostic Prediction for Non-small-Cell Lung Cancer Based on Deep Neural Network and Multimodal Data 17th

学校读者我要写书评

暂无评论

Prognostic Prediction for Non-small-Cell Lung Cancer Based o...

17th International Conference on intelligent Computing, ICIC 2021

作者： Zhang, Zhong-Si Xu, Fei Jiang, Han-Jing Chen, Zhan-Heng Key Laboratory of Image Information Processing and Intelligent Control of Education Ministry of China Institute of Artificial Intelligence School of Artificial Intelligence and Automation Huazhong University of Science and Technology Wuhan430074 China Institute of Machine Learning and Systems Biology School of Electronics and Information Engineering Tongji University No. 4800 Caoan Road Shanghai201804 China

ISBN: (纸本)9783030845315

Non-small-cell lung cancer (NSCLC) is the most common lung cancer with poor prognosis. Prognostic prediction is significant in improving the prognosis of NSCLC patients. Clinical information and multi-omics data including gene expression, miRNA, copy number variations, and DNA methylation are closely related to NSCLC prognosis. In this study, we propose a deep neural network to conduct prognostic prediction for NSCLC patients based on all five types of data. Given the high dimensional features of the omics data, past works reduce the feature dimension by regression analysis or correlation sorting algorithms. A shortcoming of these methods is that only a small number of features are considered. To overcome it, we propose a convolution neural network-based feature transformation method, which considers all features and extracts the abstract representations of the omics data. Based on the representations, we predict the five-year survival status of NSCLC patients. The results show that our method achieves a more precise prediction than previous work, with the areas under the curve improved from 0.8163 to 0.8508 and the accuracy improved from 0.7544 to 0.8096. © 2021, Springer Nature Switzerland AG.

关键词： Forecasting

Prediction of Drug-Disease Associations Based on Long Short-Term Memory Network and Gaussian Interaction Profile Kernel 15th

学校读者我要写书评

暂无评论

Prediction of Drug-Disease Associations Based on Long Short-...

15th International Conference on Bio-Inspired Computing: Theories and Applications, BIC-TA 2020

作者： Jiang, Han-Jing Wang, Yan-Bin Huang, Yabing Key Laboratory of Image Information Processing and Intelligent Control of Education Ministry of China Institute of Artificial Intelligence School of Artificial Intelligence and Automation Huazhong University of Science and Technology Wuhan430074 China College of Computer Science and Technology Zhejiang University Hangzhou315000 China Department of Pathology Renmin Hospital of Wuhan University Wuhan430060 China

ISBN: (纸本)9789811613531

Drug-disease association depicts the accessibility landscape of the association at the molecular level, thereby revealing the intermolecular reaction in the process of drug reposition. However, the scarcity of drug-disease association data often complicates the drug reposition. In this work, we present a method for predicting drug-disease association, which is called drug-disease association prediction via Gaussian interaction profile kernel and long short-term memory network (G-LSTM). G-LSTM combines a recurrent network framework with a Gaussian interaction profile kernel to learn underlying features that accurately characterize drug-disease-association data. We have verified G-LSTM models on the two benchmark datasets Fdataset and Cdataset, respectively. The experimental results show that the predictive performance was significantly improved, with AUC of 0.9527 and 0.9705, respectively. © 2021, Springer Nature Singapore Pte Ltd.

关键词： Long short-term memory

Source-free Unsupervised Domain Adaptation for Privacy-Preserving intelligent Fault Diagnosis

学校读者我要写书评

暂无评论

Source-free Unsupervised Domain Adaptation for Privacy-Prese...

Sensing, Measurement & Data Analytics in the era of Artificial Intelligence (ICSMD), International Conference on

作者： Mengliang Zhu Xiangyu Zeng Jie Liu Kaibo Zhou MOE Key Laboratory of Image Information Processing and Intelligent Control School of Artificial Huazhong University of Science and Technology Wuhan P. R. China School of Civil and Hydraulic Engineering Huazhong University of Science and Technology Wuhan P. R. China

Rolling bearing is of vital significance in industrial applications and intelligent fault diagnosis (IFD) have been widely exploited in this field. However, cross-machine variations hinder the model performance across different machines in varying working conditions. To solve this issue, many unsupervised domain adaptation (UDA) approaches have been proposed recently, requiring direct access to the fully labeled source domain. However, privacy concerns are raised in traditional UDA scenario, since raw signals contain various private information. To address this issue, the source-free unsupervised domain adaptation (SFUDA) scenario is considered, where only the pre-trained source model is required, instead of the fully labeled source domain. In this paper, a SFUDA approach for intelligent fault diagnosis (IFD) is proposed. It consists of two steps: 1) source model generalization, where virtual adversarial training, R-Drop and label smoothing techniques are adopted to improve the generalization ability of the source model; and 2) target model adaptation, where information maximization is used for cluster assumption and mean teacher training paradigm is utilized to alleviate the catastrophic forgetting phenomenon. The proposed approach is verified on the Case Western Reserve University (CWRU) dataset. Experimental results on the show that the proposed SFUDA approach outperforms several typical UDA approaches, and its effectiveness is verified.

关键词： Fault diagnosis Training Employee welfare Adaptation models Privacy Smoothing methods Data analysis

Exploiting Distilled Learning for Deep Siamese Tracking

学校读者我要写书评

暂无评论

Exploiting Distilled Learning for Deep Siamese Tracking

International Conference on Pattern Recognition

作者： Chengxin Liu Zhiguo Cao Wei Li Yang Xiao Shuaiyuan Du Angfan Zhu Key Laboratory of Image Processing and Intelligent Control Ministry of Education School of Artificial Intelligence and Automation Huazhong University of Science and Technology Wuhan China Queen Mary University of London

Existing deep siamese trackers are typically built on off-the-shelf CNN models for feature learning, with the demand for huge power consumption and memory storage. This limits current deep siamese trackers to be carried on resource-constrained devices like mobile phones, given factor that such a deployment normally requires cost-effective considerations. In this work, we address this issue by presenting a novel Distilled Learning Framework(DLF) for siamese tracking, which aims at learning tracking model with efficiency and high accuracy. Specifically, we propose two simple yet effective knowledge distillation strategies, denote as point-wise distillation and pairwise distillation, which are designed for transferring knowledge from a more discriminative teacher tracker into a compact student tracker. In this way, cost-effective and high performance tracking could be achieved. Extensive experiments on several tracking benchmarks demonstrate the effectiveness of our proposed method.

关键词： Power demand Pipelines Memory management Benchmark testing Mobile handsets Pattern recognition

A Stronger Baseline for Ego-Centric Action Detection

学校读者我要写书评

暂无评论

arXiv 2021年

作者： Qing, Zhiwu Huang, Ziyuan Wang, Xiang Feng, Yutong Zhang, Shiwei Jiang, Jianwen Tang, Mingqian Gao, Changxin Ang, Marcelo H. Sang, Nong Key Laboratory of Image Processing and Intelligent Control School of Artificial Intelligence and Automation Huazhong University of Science and Technology ARC National University of Singapore Tsinghua University Alibaba Group

This technical report analyzes an egocentric video action detection method we used in the 2021 EPIC-KITCHENS-100 competition hosted in CVPR2021 Workshop. The goal of our task is to locate the start time and the end time of the action in the long untrimmed video, and predict action category. We adopt sliding window strategy to generate proposals, which can better adapt to short-duration actions. In addition, we show that classification and proposals are conflict in the same network. The separation of the two tasks boost the detection performance with high efficiency. By simply employing these strategy, we achieved 16.10% performance on the test set of EPIC-KITCHENS-100 Action Detection challenge using a single model, surpassing the baseline method by 11.7% in terms of average mAP. Copyright © 2021, The Authors. All rights reserved.

关键词： Machine learning

Causal Invariant Hierarchical Molecular Representation for Out-of-distribution Molecular Property Prediction

学校读者我要写书评

暂无评论

Causal Invariant Hierarchical Molecular Representation for O...

IEEE International Conference on Bioinformatics and Biomedicine (BIBM)

作者： Xinlong Wen Yifei Guo Shuoying Wei Wenhan Long Lida Zhu Rongbo Zhu Huazhong Agricultural University Wuhan China College of Informatics Huazhong Agricultural University Wuhan China Shenzhen Institute of Nutrition and Health Huazhong Agricultural University Shenzhen China Shenzhen Branch Guangdong Laboratory for Lingnan Modern Agriculture Genome Analysis Laboratory of the Ministry of Agriculture and Agricultural Genomics Institute at Shenzhen Chinese Academy of Agricultural Sciences Shenzhen China Engineering Research Center of Intelligent Technology for Agriculture Ministry of Education Wuhan China Huazhong University of Science and Technology Wuhan China Key Laboratory of Ministry of Education for Image Processing and Intelligent Control School of Artificial Intelligence and Automation Huazhong University of Science and Technology Wuhan China

ISBN: (数字)9798350386226

ISBN: (纸本)9798350386233

Molecular representation learning is widely used in the field of drug discovery, due to its ability to accurately capture the complex features of compounds in high-dimensional space. However, existing molecular representation learning models are prone to be influenced by spurious parts during distribution shifts (also known as out-of-distribution, or OOD), which results in models mistakenly treating these spurious parts as crucial features of molecules, thereby limiting the generalization capability of the models. To tackle this issue, a novel invariant molecular representation learning model, called Causal Invariant Hierarchical Molecular Representation Graph Neural Networks (CHiMoGNN), is proposed for OOD molecular property prediction. In CHiMoGNN, a Feature Enhancement (FE) module is designed to leverage the multi-level molecular parts to enhance the expression of invariant features, thereby enhancing the model’s capability to capture key molecular information. In addition, a Cartesian Product based Environmental Impact (EI) module is adopted to generate counterfactual samples with environmental diversity. Consequently, these samples are utilized to train a classifier that maintains consistent performance across various environments. Extensive experiments on seven real-world datasets demonstrate that CHiMoGNN outperforms 9 state-of-the-art models, achieving a 5.73% increase in average ROC-AUC, and the results also show that CHiMoGNN can effectively maintain generalization in various distribution shifts. Code and datasets are available at https://***/Chertuion/CHiMoGNN.

关键词： Representation learning Limiting Codes Predictive models Feature extraction Iron Graph neural networks Drug discovery Data mining Compounds