检索结果-内蒙古大学图书馆

FPSMix: data augmentation strategy for point cloud classification

Frontiers of Computer Science 2025年第2期19卷 105-113页

作者： Taiyan CHEN Xianghua YING Key Lab of Machine Perception(MoE) School of Intelligence Science and TechnologyPeking UniversityBeijing 100871China

Data augmentation is a widely used regularization strategy in deep neural networks to mitigate overfitting and enhance *** the context of point cloud data,mixing two samples to generate new training examples has proven to be *** this paper,we propose a novel and effective approach called Farthest Point Sampling Mix(FPSMix)for augmenting point cloud *** method leverages farthest point sampling,a technique used in point cloud processing,to generate new samples by mixing points from two original point *** key innovation of our approach is the introduction of a significance-based loss function,which assigns weights to the soft labels of the mixed samples based on the classification loss of each part of the new sample that is separated from the two original point *** way,our method takes into account the importance of different parts of the mixed sample during the training process,allowing the model to learn better global *** results demonstrate that our FPSMix,combined with the significance-based loss function,improves the classification accuracy of point cloud models and achieves comparable performance with state-of-the-art data augmentation ***,our approach is complementary to techniques that focus on local features,and their combined use further enhances the classification accuracy of the baseline model.

关键词： point cloud classification data augmentation loss function point cloud understanding point cloud analysis

来源：评论

学校读者我要写书评

暂无评论

A Combined Approach of Vision Transformer and Transfer Learning Based Model for Accurate Lung Cancer Classification

引用

SN Computer Science 2025年第5期6卷 1-11页

作者： Kumar, Arvind Ansari, Md Khalid Singh, Koushlendra Kumar Machine Vision and Intelligence Lab National Institute of Technology Jamshedpur India

While considering the global health issue, early diagnosis of lung cancer has become a potential demand of any health-conscious system. Although the computer-aided diagnosis (CAD) system contributes a vital role in cancer-related disease they often fail to classify the different types of lung cancer due to computational complexity and high error rate. In this article, we have proposed a novel approach to classify lung cancer using transfer learning techniques and transformer-based models. The present work undertakes three major tasks: pre-processing, feature extractions, and classifications. Here, feature extraction is performed by transfer learning techniques using the VGG19 model, whereas the classification is performed with the help of the Vision Transformer (ViT) model. The combined approach of transfer learning and transformer-based model has been proven to classify lung cancer with the best accuracy of 99.80 and 99.06% over two datasets, LC25000 and IQ-OTH/NCCD, respectively. © The Author(s), under exclusive licence to Springer Nature Singapore Pte Ltd. 2025.

关键词： Classification Diagnosis LC25000 VGG19 Vision Transformer

来源：评论

学校读者我要写书评

暂无评论

SpikingMiniLM: energy-efficient spiking transformer for natural language understanding

引用

Science China(Information Sciences) 2024年第10期67卷 115-128页

作者： Jiayu ZHANG Jiangrong SHEN Zeke WANG Qinghai GUO Rui YAN Gang PAN Huajin TANG College of Computer Science and Technology Zhejiang University The State Key Lab of Brain-Machine Intelligence Zhejiang University Collaborative Innovation Center of Artificial Intelligence Zhejiang University Advanced Computing and Storage Laboratory Huawei Technologies Co. Ltd. College of Computer Science and Technology Zhejiang University of Technology MOE Frontier Science Center for Brain Science and Brain-Machine Integration Zhejiang University

In the era of large-scale pretrained models, artificial neural networks（ANNs） have excelled in natural language understanding（NLU） tasks. However, their success often necessitates substantial computational resources and energy consumption. To address this, we explore the potential of spiking neural networks（SNNs） in NLU——a promising avenue with demonstrated advantages, including reduced power consumption and improved efficiency due to their event-driven characteristics. We propose the SpikingMiniLM,a novel spiking Transformer model tailored for natural language understanding. We first introduce a multi-step encoding method to convert text embeddings into spike trains. Subsequently, we redesign the attention mechanism and residual connections to make our model operate on the pure spike-based paradigm without any normalization technique. To facilitate stable and fast convergence, we propose a general parameter initialization method grounded in the stable firing rate principle. Furthermore, we apply an ANN-to-SNN knowledge distillation to overcome the challenges of pretraining SNNs. Our approach achieves a macro-average score of 75.5 on the dev sets of the GLUE benchmark, retaining 98% of the performance exhibited by the teacher model MiniLMv2. Our smaller model also achieves similar performance to BERTMINIwith fewer parameters and much lower energy consumption, underscoring its competitiveness and resource efficiency in NLU tasks.

关键词： spiking neural networks natural language understanding spiking Transformer spike-based attention multi-step encoding ANN-to-SNN distillation

来源：评论

学校读者我要写书评

暂无评论

Study of Tokenization Strategies for the Santhali Language

引用

SN Computer Science 2024年第7期5卷 807页

作者： Ohm, Anand Kumar Singh, Koushlendra Kumar Machine Vision and Intelligence Lab Department of CSE National Institute of Technology Jamshedpur Jharkhand 831014 India

Santhali is one of the popular local languages and mainly spoken in the ruler areas of Jharkhand, Odisha, and West Bengal in Bharat. Santhali language is area specific language that is popular mainly among the tribe population. Hence, the technical aspects have not been explored. Applying different natural language processing (NLP) methods in the Santhali language is challenging. Translating the Santhali language to any other language tokenization plays a vital role, and it is the first step of language tokenization. The tokenization reduces the sentence of a paragraph to an atomic meaningful word, and it can be a single word and structure. Multiple methods are used to tokenize a language. The NLP-related tasks as popular Indian languages in India have developed technologically and are rich in resource availability. Santhali language is a resource language and does not see much use of technology. The present work encompasses the tokenization of the Santhali language. The present work has applied the four different tokenization methods to the Santhali language. It has been observed that their outputs differ from the tiniest character categories to the most general form of words. Ol-Chiki font has been used for the present study. A detailed comparison of the tokenization is presented in the present manuscript. It has been noted that almost all of the Santhali characters are composed of real-world symbols and an alphabetic writing system. spaCy model performs better than other tokenization approaches. We also validated the different tokenization methods in different paragraphs written in the Ol-Chiki script for the Santhali language. © The Author(s), under exclusive licence to Springer Nature Singapore Pte Ltd. 2024.

关键词： Low resource language machine learning NLP Ol-Chiki Santhali Tokenization

来源：评论

学校读者我要写书评

暂无评论

TAL: Two-stream Adaptive Learning for Generalizable Person Re-identification

引用

machine intelligence Research 2025年第2期22卷 337-351页

作者： Yichao Yan Junjie Li Shengcai Liao Jie Qin MoE Key Lab of Artificial Intelligence AI InstituteShanghai Jiao Tong UniversityShanghai 200240China Inception Institute of Artificial Intelligence Abu Dhabi 999041UAE College of Artificial Intelligence Nanjing University of Aeronautics and Astronauticsthe Key Laboratory of Brain-Machine Intelligence TechnologyMinistry of EducationNanjing 211106China

Domain generalizable person re-identification(reid)is a challenging task in computer vision,which aims to apply a trained reid model to unseen *** works either combine the data in all the training domains to capture domain-invariant features,or adopt a mixture of experts to investigate domain-specific *** this work,we argue that both domain-specific and domain-invariant features are crucial for improving the generalization ability of reid *** this end,we design a novel framework,which we name two-stream adaptive learning(TAL),to simultaneously model these two kinds of ***,a domain-specific stream is proposed to capture the training domain statistics with batch normalization(BN)parameters,whereas an adaptive matching layer is designed to dynamically aggregate domain-level *** the meantime,we design an adaptive BN layer in the domain-invariant stream to approximate the statistic of unseen domains,such that our model is capable of handling various novel *** two streams work adaptively and collaboratively to learn generalizable reid *** validated by extensive experiments,our framework can be applied to both single-source and multi-source domain generalization tasks,where the results show that our framework notably outperforms the state-of-the-art methods.

关键词： Person re-identification domain generalization image retrieval representation learning computer vision.

来源：评论

学校读者我要写书评

暂无评论

Tokenization's Potential for Santhali Language Processing 3rd

Tokenization's Potential for Santhali Language Processing

引用

3rd International Conference on machine Vision and Augmented intelligence, MAI 2023

作者： Ohm, Anand Kumar Panday, Aman Kumar Singh, Koushlendra Kumar Machine Vision and Intelligence Lab Department of CSE National Institute of Technology Jamshedpur Jharkhand India

ISBN: (纸本)9789819743582

Santhali is one of the popular local languages and is mainly spoken by tribal people of States like Jharkhand, Odisha, and West Bengal, as well as a few other parts of the country. Tokenization is one of the major challenges in Natural Language Processing (NLP). Tokenization reduces the length of sentences and paragraphs. There are various methods in use to tokenize different languages depending on the writing and structure of the language. Tokenization is important in many tasks like machine translation, text summarization, chatbot creation, and spam detection. In Natural Language Processing (NLP) most of the Indian spoken languages have high resource availability and have improved technologically in various fields. Still, a low-resource language like SANTHALI does not make much of its existence in technology. The present work is based on the tokenization of Santhali. All available tokenizers to tokenize in the Santhali language have been explored. Ol-Chiki font has been used for the present study. A detailed comparison of tokenization has been represented in the manuscript in the present study. It has been observed that for Santhali a writing system is alphabetic and most of the characters are made of real-world symbols. spaCy model performs better than other available tokenizer models based on the mechanism of the Santhali language. © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2025.

关键词： Natural language processing systems

来源：评论

学校读者我要写书评

暂无评论

Newton design:designing CNNs with the family of Newton's methods

引用

Science China(Information Sciences) 2023年第6期66卷 122-137页

作者： Zhengyang SHEN Yibo YANG Qi SHE Changhu WANG Jinwen MA Zhouchen LIN School of Mathematical Sciences Peking University Bytedance AI Lab JD Explore Academy Key Laboratory of Machine Perception School of Intelligence Science and TechnologyPeking University Pazhou Lab

Nowadays, convolutional neural networks(CNNs)have led the developments of machine ***, most CNN architectures are obtained by manual design, which is empirical, time-consuming, and non-transparent. In this paper, we aim at offering better insight into CNN models from the perspective of optimization theory. We propose a unified framework for understanding and designing CNN architectures with the family of Newton's methods, which is referred to as Newton design. Specifically, we observe that the standard feedforward CNN model(PlainNet)solves an optimization problem via a kind of quasi-Newton method. Interestingly, residual network(ResNet)can also be derived if we use a more general quasi-Newton method to solve this problem. Based on the above observations, we solve this problem via a better method,the Newton-conjugate-gradient(Newton-CG)method, which inspires Newton-CGNet. In the network design,we translate binary-value terms in the optimization schemes to dropout layers, so dropout modules naturally appear in the derived CNN structures with specific locations, rather than being an empirical training *** experiments on image classification and text categorization tasks verify that Newton-CGNets perform very competitively. Particularly, Newton-CGNets surpass their counterparts ResNets by over 4% on CIFAR-10 and over 10% on CIFAR-100, respectively.

关键词： CNN dropout optimization method network design Newton's method

来源：评论

学校读者我要写书评

暂无评论

An approach to classify lung and colon cancer of histopathology images using deep feature extraction and an ensemble method

引用

International Journal of Information technology (Singapore) 2023年第8期15卷 4149-4160页

作者： Singh, Onkar Singh, Koushlendra Kumar Machine Vision and Intelligence Lab Department of CSE National Institute of Technology Jamshedpur Jamshedpur India

Cancer is a deadly condition brought on by a range of metabolic anomalies and genetic disorders. Lung and colon cancer in particular is among the leading causes of death and disease in humans. The histopathological diagnosis is crucial in identifying this type of cancer. The histopathological lung cancer diagnostic test is a crucial part of the patient's treatment. The detection and classification of lung and colon cancers are thus among the most important study objectives, especially in the field of medical and biological health systems. In the current research, we proposed an ensemble classifier utilizing three distinct approaches random forest (RF), support vector machine (SVM), and logistic regression (LR) model. The predictions from each classifier are pooled by using the majority voting method to create the ensemble classifier. Deep features from lung and colon images have been extracted using two different methods: VGG16 and local binary pattern (LBP). Integrating the LBP and VGG16 features yields the initial relevant features for the ensemble classifier. The suggested methodology is validated using histopathological (LC25000) lung and colon datasets. For the identification of lung and colon cancer, our suggested technique has an average accuracy of 99.00%, precision of 99.00%, recall of 98.80%, and F_I score of 98.80%. The investigation's results indicate that our proposed method works significantly better than the existing models. When compared to employing a single classifier individually, this methodology is suitable to solve a wide variety of classification issues and may increase prediction performance. © 2023, The Author(s), under exclusive licence to Bharati Vidyapeeth's Institute of Computer Applications and Management.

关键词： Histopathology images Local binary pattern (LBP) Logistic regression Lung cancer Random forest Support vector machine (SVM) VGG16

来源：评论

学校读者我要写书评

暂无评论

A Chinese Multimodal Social Video Dataset for Controversy Detection 24

A Chinese Multimodal Social Video Dataset for Controversy De...

引用

32nd ACM International Conference on Multimedia, MM 2024

作者： Xu, Tianjiao Chen, Aoxuan Zhao, Yuxi Gao, Jinfei Gan, Tian Shandong University Qingdao China MoE Key Lab of Brain-Machine Intelligence Technology Nuaa China

ISBN: (纸本)9798400706868

Social video platforms have emerged as significant channels for information dissemination, facilitating lively public discussions that often give rise to controversies. However, existing approaches to controversy detection primarily focus on textual features, which raises three key concerns: it underutilizes the potential of visual information available on social media platforms;it is ineffective when faced with incomplete or absent textual information;and the existing datasets fail to adequately address the need for comprehensive multimodal resources on social media platforms. To address these challenges, we construct a large-scale Multimodal Controversial Dataset (MMCD) in Chinese. Additionally, we propose a novel framework named Multi-view Controversy Detection (MVCD) to effectively model controversies from multiple perspectives. Through extensive experiments using state-of-the-art models on the MMCD, we demonstrate MVCD's effectiveness and potential impact. © 2024 ACM.

关键词： controversy detection

来源：评论

学校读者我要写书评

暂无评论

FreeLong: Training-Free Long Video Generation with SpectralBlend Temporal Attention 38

FreeLong: Training-Free Long Video Generation with SpectralB...

引用

38th Conference on Neural Information Processing Systems, NeurIPS 2024

作者： Lu, Yu Liang, Yuanzhi Zhu, Linchao Yang, Yi The State Key Lab of Brain-Machine Intelligence Zhejiang University China University of Technology Sydney Australia

Video diffusion models have made substantial progress in various video generation applications. However, training models for long video generation tasks require significant computational and data resources, posing a challenge to developing long video diffusion models. This paper investigates a straightforward and training-free approach to extend an existing short video diffusion model (e.g., pre-trained on 16-frame videos) for consistent long video generation (e.g., 128 frames). Our preliminary observation has found that directly applying the short video diffusion model to generate long videos can lead to severe video quality degradation. Further investigation reveals that this degradation is primarily due to the distortion of high-frequency components in long videos, characterized by a decrease in spatial high-frequency components and an increase in temporal high-frequency components. Motivated by this, we propose a novel solution named FreeLong to balance the frequency distribution of long video features during the denoising process. FreeLong blends the low-frequency components of global video features, which encapsulate the entire video sequence, with the high-frequency components of local video features that focus on shorter subsequences of frames. This approach maintains global consistency while incorporating diverse and high-quality spatiotemporal details from local videos, enhancing both the consistency and fidelity of long video generation. We evaluated FreeLong on multiple base video diffusion models and observed significant improvements. Additionally, our method supports coherent multi-prompt generation, ensuring both visual coherence and seamless transitions between scenes. Our project page is at: https://***/freelong. © 2024 Neural information processing systems foundation. All rights reserved.

关键词：

来源：评论

学校读者我要写书评

暂无评论

建议与咨询留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

时间限定

文献类型

馆藏选择

核心期刊

语言

文献类型

帮助

文字说明：

检索规则说明：

检索范例：

分类表

所选分类

限定检索结果

文献类型

馆藏范围

日期分布

学科分类号

主题

机构

作者

语言

请选择保存的检索档案：

请选择收藏分类：

通借通还

建议与咨询 留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

时间限定

文献类型

馆藏选择

核心期刊

语言

文献类型

帮助

文字说明：

检索规则说明：

检索范例：

分类表

所选分类

限定检索结果

文献类型

馆藏范围

日期分布

学科分类号

主题

机构

作者

语言

请选择保存的检索档案： 新增检索档案 确定 取消

请选择收藏分类： 新增自定义分类 确定 取消

通借通还

建议与咨询留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

请选择保存的检索档案：

请选择收藏分类：