检索结果-内蒙古大学图书馆

Channel DropBlock: An improved regularization method for fine-grained visual classification

学校读者我要写书评

暂无评论

arXiv 2021年

作者： Ding, Yifeng Dong, Shuwei Tong, Yujun Ma, Zhanyu Xiao, Bo Ling, Haibin Pattern Recognition and Intelligent System Laboratory Beijing University of Posts and Telecommunications Department of Computer Science Stony Brook University

Classifying the sub-categories of an object from the same super-category (e.g., bird) in a fine-grained visual classification (FGVC) task highly relies on mining multiple discriminative features. Existing approaches mainly tackle this problem by introducing attention mechanisms to locate the discriminative parts or feature encoding approaches to extract the highly parameterized features in a weakly-supervised fashion. In this work, we propose a lightweight yet effective regularization method named Channel DropBlock (CDB), in combination with two alternative correlation metrics, to address this problem. The key idea is to randomly mask out a group of correlated channels during training to destruct features from co-adaptations and thus enhance feature representations. Extensive experiments on three benchmark FGVC datasets show that CDB effectively improves the performance. Copyright © 2021, The Authors. All rights reserved.

关键词： Benchmarking

Modeling discriminative representations for out-of-domain detection with supervised contrastive learning

学校读者我要写书评

暂无评论

arXiv 2021年

作者： Zeng, Zhiyuan He, Keqing Yan, Yuanmeng Liu, Zijun Wu, Yanan Xu, Hong Jiang, Huixing Xu, Weiran Pattern Recognition & Intelligent System Laboratory Beijing University of Posts and Telecommunications Beijing China Meituan Group Beijing China

Detecting Out-of-Domain (OOD) or unknown intents from user queries is essential in a task-oriented dialog system. A key challenge of OOD detection is to learn discriminative semantic features. Traditional cross-entropy loss only focuses on whether a sample is correctly classified, and does not explicitly distinguish the margins between categories. In this paper, we propose a supervised contrastive learning objective to minimize intra-class variance by pulling together in-domain intents belonging to the same class and maximize inter-class variance by pushing apart samples from different classes. Besides, we employ an adversarial augmentation mechanism to obtain pseudo diverse views of a sample in the latent space. Experiments on two public datasets prove the effectiveness of our method capturing discriminative representations for OOD detection. © 2021, CC BY.

关键词： Semantics

Face transformer for recognition

学校读者我要写书评

暂无评论

arXiv 2021年

作者： Zhong, Yaoyao Deng, Weihong Pattern Recognition and Intelligent System Laboratory School of Artificial Intelligence Beijing University of Posts and Telecommunications Beijing100876 China

Recently there has been a growing interest in Transformer not only in NLP but also in computer vision. We wonder if transformer can be used in face recognition and whether it is better than CNNs. Therefore, we investigate the performance of Transformer models in face recognition. Considering the original Transformer may neglect the interpatch information, we modify the patch generation process and make the tokens with sliding patches which overlaps with each others. The models are trained on CASIA-WebFace and MSCeleb- 1M databases, and evaluated on several mainstream benchmarks, including LFW, SLLFW, CALFW, CPLFW, TALFW, CFP-FP, AGEDB and IJB-C databases. We demonstrate that Face Transformer models trained on a large-scale database, MS-Celeb-1M, achieve comparable performance as CNN with similar number of parameters and MACs. To facilitate further researches, Face Transformer models and codes are available at https://***/zhongyy/Face-Transformer. Copyright © 2021, The Authors. All rights reserved.

关键词： Face recognition

Novel slot detection: A benchmark for discovering unknown slot types in the task-oriented dialogue system

学校读者我要写书评

暂无评论

arXiv 2021年

作者： Wu, Yanan Zeng, Zhiyuan He, Keqing Xu, Hong Yan, Yuanmeng Jiang, Huixing Xu, Weiran Pattern Recognition & Intelligent System Laboratory Beijing University of Posts and Telecommunications Beijing China Meituan Group Beijing China

Existing slot filling models can only recognize pre-defined in-domain slot types from a limited slot set. In the practical application, a reliable dialogue system should know what it does not know. In this paper, we introduce a new task, Novel Slot Detection (NSD), in the task-oriented dialogue system. NSD aims to discover unknown or out-of-domain slot types to strengthen the capability of a dialogue system based on in-domain training data. Besides, we construct two public NSD datasets, propose several strong NSD baselines, and establish a benchmark for future work. Finally, we conduct exhaustive experiments and qualitative analysis to comprehend key challenges and provide new guidance for future directions. © 2021, CC BY.

关键词： Speech processing

IEEE T-BIOM Editorial Board Changes

学校读者我要写书评

暂无评论

IEEE Transactions on Biometrics, Behavior, and Identity Scie...

IEEE Transactions on Biometrics, Behavior, and Identity Science 2024年第1期7卷 1-2页

作者： Naser Damer Weihong Deng Jianjiang Feng Vishal M. Patel Ajita Rattani Mark Nixon Smart Living & Biometric Technologies Fraunhofer Institute for Computer Graphics Research IGD Darmstadt Germany Pattern Recognition and Intelligent System Laboratory Beijing University of Posts and Telecommunications Beijing China Department of Automation Tsinghua University Beijing China Department of Electrical and Computer Engineering Johns Hopkins University Baltimore MD USA Department of Computer Science and Engineering College of Engineering University of North Texas Denton TX USA

Caption Feature Space Regularization for Audio Captioning

学校读者我要写书评

暂无评论

arXiv 2022年

作者： Zhang, Yiming Yu, Hong Du, Ruoyi Ma, Zhanyu Dong, Yuan The Pattern Recognition and Intelligent System Laboratory School of Artificial Intelligence Beijing University of Posts and Telecommunications Beijing100876 China Department of Artificial Intelligence School of Information and Electrical Engineering Ludong University Shandong Yantai264025 China

Audio captioning aims at describing the content of audio clips with human language. Due to the ambiguity of audio, different people may perceive the same audio differently, resulting in caption disparities (i.e., one audio may correlate to several captions with diverse semantics). For that, general audio captioning models achieve the one-to-many training by randomly selecting a correlated caption as the ground truth for each audio. However, it leads to a significant variation in the optimization directions and weakens the model stability. To eliminate this negative effect, in this paper, we propose a two-stage framework for audio captioning: (i) in the first stage, via the contrastive learning, we construct a proxy feature space to reduce the distances between captions correlated to the same audio, and (ii) in the second stage, the proxy feature space is utilized as additional supervision to encourage the model to be optimized in the direction that benefits all the correlated captions. We conducted extensive experiments on two datasets using four commonly used encoder and decoder architectures. Experimental results demonstrate the effectiveness of the proposed method. The code is available at https://github. com/PRIS-CV/Caption-Feature-Space-Regularization. © 2022, CC BY.

关键词： Semantics

Feature Decoupled of Deep Mutual Information Maximization

学校读者我要写书评

暂无评论

Feature Decoupled of Deep Mutual Information Maximization

Automation, Robotics and Computer Engineering (ICARCE), International Conference on

作者： Xing He Changgen Peng Lin Wang Weijie Tan Zifan Wang State Key Laboratory of Public Big Data College of Computer Science and Technology Guizhou University Guiyang China Guizhou Big Data Academy Guizhou University Guiyang China Guizhou Key Laboratory of Pattern Recognition and Intelligent System Guizhou Minzu University Guiyang China Institute of Guizhou Aerospace Measuring and Testing Technology Guiyang China

In deep learning, supervised learning techniques usually require a large amount of expensive labeled data to train the network, and the feature representations extracted by the model usually mix multiple attributes, resulting in feature representations that are difficult to decouple and are non-interpretable, which restricts the application and development of deep learning techniques, and for this reason, it is particularly important to study decoupled feature representation methods for unsupervised learning. Although the Learning deep representations by mutual information estimation and maximization (DIM) method achieves excellent results in unsupervised learning, the feature representations learned by the DIM method still suffer from the problem of difficult decoupling. decoupling problem. To address this problem, we minimize the mutual information between the intermediate layer feature representations learned by the hidden layer of the encoder during the encoder training process, so that the features learned by each filter are as uncorrelated as possible, thus realizing feature decoupling, and our method is called FP-DIM. Finally, the effectiveness of the FP-DIM method is verified on the CIFAR-10, STL-10, and Fashion-MNIST datasets. The experiments show that our proposed FP-DIM method is more significant for the learned decouplable middle layer feature representation. Finally, we also propose a reflection of future research for the FP-DIM method, aiming to provide a research idea and direction for solving unsupervised interpretable machine learning and to lay a solid theoretical and application foundation for machine learning fields such as image classification and migration learning.

关键词：

Fine-Grained Visual Classification via Progressive Multi-granularity Training of Jigsaw Patches 1

学校读者我要写书评

暂无评论

16th European Conference on Computer Vision, ECCV 2020

作者： Du, Ruoyi Chang, Dongliang Bhunia, Ayan Kumar Xie, Jiyang Ma, Zhanyu Song, Yi-Zhe Guo, Jun Pattern Recognition and Intelligent System Laboratory School of Artificial Intelligence Beijing University of Posts and Telecommunications Beijing China SketchX CVSSP University of Surrey Guildford United Kingdom

ISBN: (数字)9783030585655

ISBN: (纸本)9783030585648

Fine-grained visual classification (FGVC) is much more challenging than traditional classification tasks due to the inherently subtle intra-class object variations. Recent works are mainly part-driven (either explicitly or implicitly), with the assumption that fine-grained information naturally rests within the parts. In this paper, we take a different stance, and show that part operations are not strictly necessary – the key lies with encouraging the network to learn at different granularities and progressively fusing multi-granularity features together. In particular, we propose: (i) a progressive training strategy that effectively fuses features from different granularities, and (ii) a random jigsaw patch generator that encourages the network to learn features at specific granularities. We evaluate on several standard FGVC benchmark datasets, and show the proposed method consistently outperforms existing alternatives or delivers competitive results. The code is available at https://***/PRIS-CV/PMG-Progressive-Multi-Granularity-Training. © 2020, Springer Nature Switzerland AG.

关键词： Computer vision

Vehicle Routing Problem with Time Windows using Hybrid Metaheuristic Dragonfly Algorithm and Variable Neighborhood Search: Work on Progress

学校读者我要写书评

暂无评论

Vehicle Routing Problem with Time Windows using Hybrid Metah...

International Conference on Electrical Engineering, Computer Science and Informatics (EECSI)

作者： Yunita Yunita Deris Stiawan Dian Palupi Rini Doctoral Student at Faculty of Computer Science Sriwijaya University Intelligent System Research Group Faculty of Computer Science Sriwijaya University Palembang South Sumatera Communication Network and Information Security Research Group Faculty of Computer Science Sriwijaya University Palembang South Sumatera Image Processing dan Pattern Recognition Laboratory Group Faculty of Computer Science Sriwijaya University Palembang South Sumatera

ISBN: (数字)9798350355314

ISBN: (纸本)9798350355321

One of the problems with Smart Transportation is the problem of cost and travel time. This problem is known as the Variable Routing Problem (VRP). In some real cases, in addition to considering route selection, there are limitations on the capacity and time period for the vehicle to serve each customer known as Time Windows, so VRP has developed into Variable Routing Problem with Time Windows (VRPTW). One way to solve VRPTW is to use metaheuristic methods. However, the metaheuristic method has a weakness, which are it can get stuck in local optimums and fail to find better global solutions. To overcome this, a combination with other methods or known as hybrid is needed. The formulation of this research problem is how to develop a Smart transportation model using a metaheuristic hybrid algorithm. The purpose of this study is to develop a Smart transportation model using a metaheuristic hybrid algorithm. The method used is to combine two metaheuristic algorithms, which are the Dragonfly Algorithm (DA) and the Variable Neighborhood Search (VNS). The result of this research is a new algorithm model which is a hybrid between DA and VNS. Through a combination of global exploration using DA and local exploitation using VNS, this algorithm is expected to be able to find a better solution and faster convergence in solving VRPTW problems

关键词： Electrical engineering Computer science Costs Metaheuristics Vehicle routing Routing Search problems Smart transportation Informatics Convergence