Article
Computer Science, Artificial Intelligence
Min Yuan, Yitian Xu
Summary: This research introduces the safe screening rule to mitigate the storage burden of multiple-instance SVM. It designs an instance elimination strategy, a dual screening method, and a smart dual coordinate descent method for faster solution efficiency.
APPLIED SOFT COMPUTING
(2022)
Article
Engineering, Industrial
Haijuan Cui, Xiaochuan Luo, Yuan Wang
Summary: This paper addresses a steelmaking-continuous casting scheduling problem by introducing Lagrangian relaxation and surrogate subgradient algorithm. The improved concave-convex procedure is designed to decompose the problem into tractable subproblems, with analysis on convergence. Computational experiments demonstrate the effectiveness of the proposed methods.
INTERNATIONAL JOURNAL OF PRODUCTION RESEARCH
(2022)
Article
Computer Science, Artificial Intelligence
Haiyan Chen, Yizhen Jia, Jiaming Ge, Bin Gu
Summary: Ordinal regression is a popular method for solving multi-class problems with ranked samples. Semi-supervised ordinal regression, which utilizes unlabeled samples, is important for data mining applications. This paper proposes an incremental learning algorithm IL-SSOR for SSOR and analyzes its convergence.
Article
Computer Science, Artificial Intelligence
Xi Wang, Fangyao Tang, Hao Chen, Carol Y. Cheung, Pheng-Ann Heng
Summary: Supervised deep learning has achieved success in DME recognition from OCT volumetric images. However, the shortage of labeled data and the expensive annotation make accurate analysis difficult. To tackle this problem, the authors propose a deep semi-supervised multiple instance learning framework that leverages a small amount of labeled data and a large amount of unlabeled data.
MEDICAL IMAGE ANALYSIS
(2023)
Article
Computer Science, Artificial Intelligence
Yu-Feng Li, Lan-Zhe Guo, Zhi-Hua Zhou
Summary: This paper investigates safe weakly supervised learning, aims to derive safe predictions by integrating multiple weakly supervised learners. A generic ensemble learning scheme is presented to optimize the worst-case performance gain, bringing multiple advantages to safe weakly supervised learning, demonstrated through extensive experiments on various tasks.
IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE
(2021)
Article
Computer Science, Information Systems
Toon Vanderschueren, Tim Verdonck, Bart Baesens, Wouter Verbeke
Summary: Predictive models are increasingly used to optimize decision-making and minimize costs. This work compared the predict-then-optimize approach with the predict-and-optimize approach in cost-sensitive classification. The key finding was that the decision-making strategy was generally more effective than training with a task-specific loss or their combination.
INFORMATION SCIENCES
(2022)
Article
Computer Science, Artificial Intelligence
Lingyun Song, Jun Liu, Mingxuan Sun, Xuequn Shang
Summary: This paper introduces the weakly supervised group mask network (WSGMN), which leverages the relations among regions to generate community instances with context information, robust to object variations. It generates masks for each label group and dynamically selects the most useful community instances' feature information for object recognition. Extensive experiments demonstrate the effectiveness of WSGMN in weakly supervised object detection tasks.
INTERNATIONAL JOURNAL OF COMPUTER VISION
(2021)
Article
Computer Science, Artificial Intelligence
Anusha Aswath, Ahmad Alsahaf, Ben N. G. Giepmans, George Azzopardi
Summary: This review summarizes the progress of deep learning-based segmentation techniques in large-scale cellular electron microscopy (EM) over the past six years. It discusses the application of deep learning in EM segmentation, including supervised, unsupervised, and self-supervised learning methods, and examines their adaptability in segmenting cellular and sub-cellular structures. Evaluation measures for benchmarking EM datasets in various segmentation tasks are also provided. Finally, the current trends and future prospects of EM segmentation with large-scale models and unlabeled images are discussed.
MEDICAL IMAGE ANALYSIS
(2023)
Article
Computer Science, Artificial Intelligence
Xin Huang, Qianshu Zhu, Yongtuo Liu, Shengfeng He
Summary: This paper introduces an Instance-aware Cue propagation Network (ICN) with a new proposal-matching strategy to address the challenges of using response maps for segmentation. Experimental results demonstrate that our method outperforms weakly-supervised state-of-the-arts on both semantic and instance segmentation tasks in the PASCAL VOC 2012 dataset.
Article
Computer Science, Artificial Intelligence
Yoni Schirris, Efstratios Gavves, Iris Nederlof, Hugo Mark Horlings, Jonas Teuwen
Summary: This study proposes a deep learning-based weak label learning method for analyzing tumor tissue whole slide images without pixel-level or tile-level annotations. They utilize self-supervised pre-training and heterogeneity-aware deep multiple instance learning. The method is applied to the prediction of homologous recombination deficiency and microsatellite instability, achieving improved performance.
MEDICAL IMAGE ANALYSIS
(2022)
Article
Computer Science, Artificial Intelligence
Namyup Kim, Sehyun Hwang, Suha Kwak
Summary: This paper introduces the first attempt to learn semantic boundary detection using image-level class labels as supervision. It uses an image classification network to estimate the coarse areas of object classes and formulates the task as a multiple instance learning problem. The authors also design a new neural network architecture that can learn to estimate semantic boundaries reliably. The final model trained with pseudo labels achieves outstanding performance on the SBD dataset.
INTERNATIONAL JOURNAL OF COMPUTER VISION
(2022)
Review
Biochemical Research Methods
Mohamed Nadif, Francois Role
Summary: Biomedical scientific literature is growing rapidly, making it challenging to identify relevant results; automated information extraction tools based on text mining techniques are essential; deep neural networks have significantly advanced this research field.
BRIEFINGS IN BIOINFORMATICS
(2021)
Article
Computer Science, Artificial Intelligence
Wei Gao, Fang Wan, Jun Yue, Songcen Xu, Qixiang Ye
Summary: D-MIL introduces discrepantly collaborative modules into MIL, creating complementary solutions for precise object localization through multiple MIL learners. The teachers-students model improves performance by providing rich information and absorbing complementary knowledge from multiple teachers. D-MIL achieves state-of-the-art performance on the challenging MS-COCO object detection benchmark.
PATTERN RECOGNITION
(2022)
Article
Computer Science, Artificial Intelligence
Yuanhai Lv, Chen Jiao, Wanqing Zhao, Wei Zhao, Ziyu Guan, Xiaofei He
Summary: This work proposes a hierarchical weakly supervised multi-instance hash learning method with a spatial pyramidal structure, which combines the advantages of local and multi-scale perception on CNN with the global field of view on Transformer. It leverages the principle of multi-instance learning to implement instance-level hash mapping capability in a weakly supervised learning manner. Experimental results on three public datasets show improved performance compared to traditional methods, validating the effectiveness of the proposed method.
INTERNATIONAL JOURNAL OF MACHINE LEARNING AND CYBERNETICS
(2023)
Article
Computer Science, Information Systems
Longrong Yang, Hongliang Li, Qingbo Wu, Fanman Meng, Heqian Qiu, Linfeng Xu
Summary: This paper proposes a bias-correction feature learner for addressing the issue of model learning bias in instance segmentation. By discarding incorrect supervision and constructing high-quality positive pairs, the method is able to accurately extract foreground features.
IEEE TRANSACTIONS ON MULTIMEDIA
(2023)
Article
Computer Science, Artificial Intelligence
Guan'an Wang, Yang Yang, Tianzhu Zhang, Jian Cheng, Zengguang Hou, Prayag Tiwari, Hari Mohan Pandey
Article
Computer Science, Information Systems
Yaoyu Li, Hantao Yao, Tianzhu Zhang, Changsheng Xu
Summary: PSRL proposes a novel method to improve the descriptive ability of person representation by fusing local features considering the person structure. The architecture includes two important modules: Local Semantic Feature Extraction and Structured Person Representation Learning.
ACM TRANSACTIONS ON MULTIMEDIA COMPUTING COMMUNICATIONS AND APPLICATIONS
(2021)
Editorial Material
Engineering, Electrical & Electronic
Hongliang Li, Lu Fang, Tianzhu Zhang
IEEE TRANSACTIONS ON CIRCUITS AND SYSTEMS FOR VIDEO TECHNOLOGY
(2020)
Article
Computer Science, Information Systems
Xinhong Ma, Tianzhu Zhang, Changsheng Xu
IEEE TRANSACTIONS ON MULTIMEDIA
(2020)
Article
Computer Science, Artificial Intelligence
Meng Meng, Tianzhu Zhang, Wenfei Yang, Jian Zhao, Yongdong Zhang, Feng Wu
Summary: Weakly Supervised Object Localization (WSOL) aims to localize objects using only image-level labels, providing better scalability and practicality than fully supervised methods. However, current techniques based on classification networks only highlight discriminative parts of objects, neglecting the entire object. To address this issue, this paper proposes a novel end-to-end part discovery model (PDM) that learns multiple discriminative object parts for accurate localization and classification.
IEEE TRANSACTIONS ON IMAGE PROCESSING
(2022)
Article
Computer Science, Information Systems
Chunxiao Liu, Zhendong Mao, Tianzhu Zhang, An-An Liu, Bin Wang, Yongdong Zhang
Summary: The paper introduces a novel focal attention mechanism to achieve more accurate semantic alignment in multimodal learning, by selectively attending to relevant sub-elements and preventing interference from irrelevant ones. Extensive experiments on image-text matching and text-to-image generation demonstrate that the focal attention significantly outperforms existing methods, providing effectiveness validation in various multimodal tasks.
IEEE TRANSACTIONS ON MULTIMEDIA
(2022)
Article
Computer Science, Artificial Intelligence
Junyu Gao, Tianzhu Zhang, Changsheng Xu
Summary: This study proposes a task-driven message passing process using a prototype-sample GNN to achieve zero-shot learning in video classification, successfully establishing relationships between categories and attributes, and achieving favorable performance on five popular video benchmarks.
IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE
(2021)
Article
Automation & Control Systems
Yehui Yang, Fangxin Shang, Binghong Wu, Dalu Yang, Lei Wang, Yanwu Xu, Wensheng Zhang, Tianzhu Zhang
Summary: This article introduces a robust framework for DR severity grading that collaboratively utilizes patch-level and image-level annotations, exchanging grade information bidirectionally to incorporate fine-grained lesion details and image-level grades for improved performance. The algorithm has shown better performance than state-of-the-art algorithms and clinical ophthalmologists, proving its robustness in facing real-world variations. Extensive ablation studies have been conducted to validate the effectiveness and necessity of each motivation in the proposed framework.
IEEE TRANSACTIONS ON CYBERNETICS
(2022)
Proceedings Paper
Computer Science, Artificial Intelligence
Wenfei Yang, Tianzhu Zhang, Xiaoyuan Yu, Tian Qi, Yongdong Zhang, Feng Wu
Summary: The proposed Uncertainty Guided Collaborative Training (UGCT) strategy effectively improves the performance of attention based methods for weakly supervised temporal action detection by generating pseudo labels online and mitigating noise in the generated labels. Experimental results show a significant performance improvement of more than 4% for all three methods on the THUMOS14 dataset.
2021 IEEE/CVF CONFERENCE ON COMPUTER VISION AND PATTERN RECOGNITION, CVPR 2021
(2021)
Proceedings Paper
Computer Science, Artificial Intelligence
Rui Sun, Yihao Li, Tianzhu Zhang, Zhendong Mao, Feng Wu, Yongdong Zhang
Summary: The study proposed a novel lesion-aware transformer (LAT) for diabetic retinopathy (DR) grading and lesion discovery, achieving the tasks through an encoder-decoder structure. This method effectively addresses the issues of lesion localization and diversity recognition, and demonstrates superior performance on multiple benchmark tests.
2021 IEEE/CVF CONFERENCE ON COMPUTER VISION AND PATTERN RECOGNITION, CVPR 2021
(2021)
Proceedings Paper
Computer Science, Artificial Intelligence
Wang Luo, Tianzhu Zhang, Wenfei Yang, Jingen Liu, Tao Mei, Feng Wu, Yongdong Zhang
Summary: This paper introduces an Action Unit Memory Network (AUMN) for weakly supervised temporal action localization, which mitigates challenges by learning action unit memory bank and utilizes diverse mechanisms. It is the first to explicitly model action units with a memory network, showing superior performance compared to state-of-the-art methods on standard benchmarks.
2021 IEEE/CVF CONFERENCE ON COMPUTER VISION AND PATTERN RECOGNITION, CVPR 2021
(2021)
Article
Computer Science, Artificial Intelligence
Wenfei Yang, Tianzhu Zhang, Zhendong Mao, Yongdong Zhang, Qi Tian, Feng Wu
Summary: This paper proposed an end-to-end Multi-Scale Structure-Aware Network (MSA-Net) for weakly supervised temporal action detection, which explores both the global and local structure information to effectively learn discriminative structure aware representations for robust and complete action detection. Extensive experimental results on two benchmark datasets demonstrate that MSA-Net outperforms state-of-the-art methods.
IEEE TRANSACTIONS ON IMAGE PROCESSING
(2021)
Article
Computer Science, Artificial Intelligence
Jianfeng He, Tianzhu Zhang, Yuhui Zheng, Mingliang Xu, Yongdong Zhang, Feng Wu
Summary: This paper proposes a novel end-to-end Consistency Graph Modeling Network (CGMNet) for semantic correspondence by jointly modeling inter-image relationship, intra-image relationship and cycle consistency. CGMNet performs well in experiments and is validated on multiple challenging datasets.
IEEE TRANSACTIONS ON IMAGE PROCESSING
(2021)
Article
Computer Science, Artificial Intelligence
Wenfei Yang, Tianzhu Zhang, Yongdong Zhang, Feng Wu
Summary: LCNet utilizes hierarchical representation of video and text features and introduces a self-supervised cycle-consistent loss to effectively learn the matching relationships between video and text, achieving superior performance compared to existing weakly supervised methods.
IEEE TRANSACTIONS ON IMAGE PROCESSING
(2021)
Article
Computer Science, Information Systems
Xiaoheng Jiang, Li Zhang, Tianzhu Zhang, Pei Lv, Bing Zhou, Yanwei Pang, Mingliang Xu, Changsheng Xu
Summary: In this study, a novel density-aware convolutional neural network (DensityCNN) method is proposed to perform crowd counting by learning density-level classification and density map estimation. Extensive experiments demonstrate the high effectiveness of the proposed method across multiple datasets.
IEEE TRANSACTIONS ON MULTIMEDIA
(2021)