检索结果-内蒙古大学图书馆

DDMCB: Open-world object detection empowered by Denoising Diffusion Models and Calibration Balance

IMAGE AND VISION COMPUTING 2025年 157卷

作者： Huang, Yangyang Xi, Xing Luo, Ronghua South China Univ Technol Guangzhou 510006 Peoples R China

Open-world object detection (OWOD) differs from traditional object detection by being more suited to real- world, dynamic scenarios. It aims to recognize unseen objects and have the skill to learn incrementally based on newly introduced knowledge. However, the current OWOD usually relies on supervising of known objects in identifying unknown objects, using high objectness scores as critical indicators of potential unknown objects. While these methods can detect unknown objects with features similar to known objects, they also classify regions dissimilar to known objects as background, leading to label bias issues. To address this problem, we leverage the knowledge from large visual models to provide auxiliary supervision for unknown objects. Additionally, we apply the Denoising Diffusion Probabilistic Model (DDPM) in OWOD scenarios. We propose an unsupervised modeling approach based on DDPM, which significantly improves the accuracy of unknown object detection. Despite this, the classifier trained during the model training process only encounters known classes, resulting in higher confidence for known classes during inference;thus, bias issues again occur. Therefore, we propose a probability calibration technique for post-processing predictions during inference. The calibration aims to reduce the probabilities of known objects and increase the probabilities of unknown objects, thereby balancing the final probability predictions. Our experiments demonstrate that the proposed method achieves significant improvements on OWOD benchmarks, with an unknown objects detection recall rate of 54.7 U-Recall, surpassing the current state-of-the-art (SOTA) methods by 44.3%. In terms of real-time performance, Our model uses a few parameters, and pure convolutional neural networks instead of intensive attention mechanisms, achieving an inference speed of 35.04 FPS, exceeding the SOTA OWOD methods based on Faster R-CNN and Deformable DETR by 2.79 and 10.95 FPS, respectively.

关键词： Open world DDPM Unsupervised Incremental learning object detection

来源：评论

学校读者我要写书评

暂无评论

SCDA-Net: Structure Completion and Density Awareness Network for LiDAR-Based 3D object detection

引用

IEEE ROBOTICS AND AUTOMATION LETTERS 2025年第5期10卷 4268-4275页

作者： Wu, Shuwen Yang, Jinfu Ma, Jiaqi Zhang, Shaochen Hao, Tianhao Li, Mingai Beijing Univ Technol Sch Informat Sci & Technol Beijing 100124 Peoples R China

As a fundamental task in various application scenarios, including autonomous driving and mobile robotic systems, 3D object detection has received extensive attention from researchers in both academia and industry. However, due to the working principle of LiDAR and external factors such as occlusion, the collected point cloud of the object is usually sparse and incomplete, which affects the performance of 3D object detector. In this letter, a Structure Completion and Density Awareness Network (SCDA-Net) is proposed for 3D object detection from point clouds. Specifically, a structure completion module is designed to predict dense shapes of complete point clouds by leveraging sequence transduction ability of the transformer architecture. Furthermore, we propose a density-aware voxel RoI pooling strategy to introduce density features that reflect the state information of the original objects in refinement stage. By restoring the complete structure of the objects and considering the true distribution of the points in raw point cloud, the proposed method achieves more accurate feature extraction and scene perception. Extensive experimental results on the KITTI and Waymo datasets demonstrate the effectiveness of the proposed SCDA-Net.

关键词： Three-dimensional displays Point cloud compression Proposals Feature extraction object detection Shape Laser radar Detectors Transformers Convolution segmentation and categorization deep learning for visual perception deep learning methods

来源：评论

学校读者我要写书评

暂无评论

Missing object detection and Tracking from Video Surveillance Camera Footage Based on Deep Learning Through Convolutional Gated Recurrent Neural Networks (CGRNN)

引用

JOURNAL OF ELECTRICAL ENGINEERING & TECHNOLOGY 2025年 1-9页

作者： Khan, A. Ayub Sabeenian, R. S. Sona Coll Technol Salem Dept ECE Salem Tamil Nadu India

Deep learning-based object detection is an important application of deep learning methods in visual understanding, video surveillance, robotics, autonomous driving systems, etc. Its outstanding features are powerful feature learning and representation capabilities compared to traditional object detection methods. To solve the problems proposed, the method, missing object detection in Surveillance Videos Based on Convolutional Gated Recurrent Neural Networks (CGRNN). In short, this research makes tracking applications effective and efficient. First, we obtained a pre-trained model using video frame images as input to Median filters. We generated 64*64 pixel images for missing object detection using the converted video dataset used in the pre-trained model. The pre-processing steps are utilised for reducing the quality of the image. The third step is to extract the colours of video images using the colour map values to estimate that this pixel will be a relative feature weight based on the Relative Scalar Color Histogram (RSCH). The fourth step is Feature extraction is to find out the features and match the image feature from one video frame object based on the maximum threshold weights using Convolutional Neural Network (CNN) with Maximum Standard External Regions (MSER) used for background features extracting. It's applied to improve features quality analysis and Feature support rate. Before classification, devoted to training the features with soft-max logical activation function with Convolutional Gated Recurrent Neural Networks (CGRNN) to classify the missing object from the video using images collected through video type dataset to train the images into the deep neural network. This results in higher classification accuracy, Precision, recall under F-values, and lower error rates. This enables a more achieved classification than other existing methods.

关键词： Deep learning Surveillance videos object detection Images Feature weights Convolutional gated recurrent neural networks (CGRNN) Color histogram Detecting images

来源：评论

学校读者我要写书评

暂无评论

FDI-YOLO: Feature disentanglement and interaction network based on YOLO for SAR object detection

引用

EXPERT SYSTEMS WITH APPLICATIONS 2025年 260卷

作者： Wang, Peng Luo, Yuan Zhu, Zhilin Shandong Technol & Business Univ Sch Informat & Elect Engn Yantai 264005 Shandong Peoples R China Yantai Inst Sci & Technol Yantai 265699 Shandong Peoples R China

Synthetic Aperture Radar (SAR) object detection is one of the key measures to ensure maritime traffic and safety. However, SAR images contain a large amount of speckle noise, which poses a challenge to traditional deep learning methods for feature extraction and processing. Therefore, we propose a YOLO-based feature disentanglement and interaction network for SAR object detection (FDI-YOLO). First, FDI-YOLO proposes a reversible cross stage partial network (RCSPNet) as the backbone. The RCSPNet uses reversible transformations to retain more complete feature information for feature extraction and decompose it into feature maps of different dimensions. Then, we propose a structure with cross-scale depth feature interaction (CDFI), which captures the local texture and global semantic information of the in-scale features using crossover frequency semantic perception (CFSP), and then strengthens the linking of the cross-scale features through bidirectional information interaction. Finally, we use an adaptive object detection head and a bounding box regression loss with a dynamic focusing mechanism to further improve the detection capability of FDI-YOLO for SAR images. We conducted experiments on three publicly available SAR datasets, SSDD, ISDD, and HRSID. On these datasets, we achieve F1 scores of 98.1%/88.6%/88.5%, AP50 scores of 98.7%/90.3%/90.9%, and AP50-95 scores of 71.0%/42.4%/64.3%, respectively. The experimental results show that FDI-YOLO is able to perform the task of SAR object detection well with less computational resources.

关键词： object detection SAR YOLOv8 Reversible column network Attention mechanism

来源：评论

学校读者我要写书评

暂无评论

A hybrid heuristic model for video surveillance-based on object detection and video anomaly classification using optimized Bi-LSTM-RBF

引用

SIGNAL IMAGE AND VIDEO PROCESSING 2025年第6期19卷 1-14页

作者： Ramasamy, Dhivya Praba Kanagaraj, Kavitha Kumarasamy, Jasmine Kumaraguru Coll Technol Elect & Commun Coimbatore 641049 Tamil Nadu India

Nowadays, video monitoring applications have become significant for observing human activity with the help of computer vision-based approaches for investigating numerous video sequences. The major goal of anomaly identification is to discover the abnormalities automatically in a short interval of time. Performing efficient anomaly detection in a video monitoring system is considered as a complex task due to video noise, spilling, and anomalies. Various anomaly detection models based on Artificial Intelligence (AI) have been developed for video surveillance;however, these models often address only specific issues and do not consider the evaluation concerns over time. Hence, this paper aims to implement a video anomaly detection model through surveillance cameras for reducing abnormal activities that enhance the security of the environment. At first, the input videos are collected from the standard benchmark datasets. These collected videos are given in the frame extraction phase. Further, the extracted frames are fed to the object detection phase, where the YOLO-V3 technique is used. Parameter optimization of YOLO-V3 is achieved using the Modified Cat and Mouse Optimization (MCMO) algorithm to improve detection performance. The object-detected frames are fed as input to the ResNet for extracting the deep features. The extracted deep features are utilized for the classification phase, where the Optimized Bi-directional Long Short Term Memory (Bi-LSTM)-Radial Basis Function (RBF) (OBi-LSTM-RBF) provides the classified anomaly outcome. The variables are optimized using the enhanced CMO algorithm for enhancing the efficacy of the anomaly classification. Simulation evaluations are carried out to reveal the effectiveness of the offered approach with diverse baseline algorithms using diverse performance measures. The offered approach shows significant enhancement in accuracy over baseline approaches. Specifically, it outperforms conventional CNNs by 45.47%, DNNs by 90.03%,

关键词： Video anomaly classification object detection Video surveillance Improved YOLO-V3 Modified cat and mouse optimization ResNet Optimized bi-directional long short term memory-radial basis function

来源：评论

学校读者我要写书评

暂无评论

YOLO-SSP: an object detection model based on pyramid spatial attention and improved downsampling strategy for remote sensing images

引用

VISUAL COMPUTER 2025年第3期41卷 1467-1484页

作者： Liu, Yongli Yang, Degang Song, Tingting Ye, Yichen Zhang, Xin Chongqing Normal Univ Coll Comp & Informat Sci Chongqing 401331 Peoples R China Chongqing Engn Res Ctr Educ Big Data Intelligent Percept & Applicat Chongqing 401331 Peoples R China Southwest Univ Coll Elect & Informat Engn Chongqing 400715 Peoples R China

object detection is an essential task in remote sensing image processing. However, the remote sensing images are characterized by large range of object sizes and complex object backgrounds, which results in challenges in the object detection task. Moreover, the detection effect of existing object detectors on remote sensing images is still not satisfactory. In order to tackle the above problems, an object detection model named YOLO-SSP for remote sensing images is proposed based on the YOLOv8m model in this paper. To begin with, the original downsampling layers are replaced with the proposed lightweight SPD-Conv module, which performs downsampling without loss of fine-grained information and improves the ability of the network to learn the feature representation. In addition, to adapt the large number of small objects in remote sensing images, a small object detection layer is added and achieves the expected results. Finally, a pyramid spatial attention mechanism is proposed to obtain the weights of different spatial positions through hierarchical pooling operations. It effectively improves the detection performance of small objects and those with complex backgrounds. We conducted ablation experiments on the DIOR dataset and compared the YOLO-SSP model with other state-of-the-art models. YOLO-SSP obtains 64.7% of mAP, which is an improvement of 2.3% relative to the baseline model. To demonstrate the generalizability and robustness of the improved model, the comparison experiments are also performed on the TGRS-HRRSD dataset and SIMD dataset with mAP of 77.2 and 64.9%, respectively. The code will be available at https://***/YongliLiu/SSP.

关键词： object detection Remote sensing images Small object Attention mechanism

来源：评论

学校读者我要写书评

暂无评论

Scale-Aware Pruning Framework for Remote Sensing object detection via Multifeature Representation

引用

IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING 2025年 63卷

作者： Hu, Zhuping Gong, Maoguo Zhao, Yue Zhang, Mingyang Lu, Yiheng Li, Jianzhao Pu, Yan Wang, Zhao Xidian Univ Sch Elect Engn Key Lab Collaborat Intelligence Syst Minist Educ Xian 710071 Peoples R China Inner Mongolia Normal Univ Acad Artificial Intelligence Coll Math Sci Hohhot 010028 Peoples R China Xidian Univ Sch Elect Engn Xian 710071 Peoples R China Xidian Univ Key Lab Collaborat Intelligence Syst Minist Educ Xian 710071 Peoples R China Xidian Univ Sch Comp Sci & Technol Xian 710071 Peoples R China

With the rapid advancements in computer vision, high-resolution remote sensing imagery has become a crucial data source for object detection. Nevertheless, effectively utilizing limited computational resources and reducing the burden on satellite edge devices remains a significant challenge. To effectively reduce model complexity while maintaining its representational capacity, this article proposes a scale-aware pruning framework (SAPF) to enhance remote sensing object detection ability. First, this article classifies the convolutional layers in object detection models into two categories: layers with a single-scale feature representation and layers with a multiscale feature representation. For convolutional layers with single-scale features, we utilize singular value decomposition (SVD) to quantify feature importance and assess filter redundancy to enhance model efficiency. By removing less critical filters, this pruning criteria aims to reduce the model size and computational load without compromising performance. However, convolutional layers with multiscale features are crucial for optimizing feature extraction and balancing information capture across various scales. To address this, this article evaluates the similarity between convolutional layers with different scales to determine the contribution of various scale features in multiscale fusion. Surprisingly, the SAPF can reduce the FLOPs and parameters, as well as ensure the representational ability obviously when the YOLO v5s and Faster-RCNN are adopted to classify the NWPU VHR-10, RSOD, and SIMD datasets. This means we can save the training computation resources for the model. Additionally, SAPF can significantly improve the efficiency of the model in object detection to ensure its real-time performance.

关键词： Multifeature representation object detection remote sensing remote sensing scale-aware pruning scale-aware pruning scale-aware pruning

来源：评论

学校读者我要写书评

暂无评论

ESOD: Efficient Small object detection on High-Resolution Images

引用

IEEE TRANSACTIONS ON IMAGE PROCESSING 2025年 34卷 183-195页

作者： Liu, Kai Fu, Zhihang Jin, Sheng Chen, Ze Zhou, Fan Jiang, Rongxin Chen, Yaowu Ye, Jieping Zhejiang Univ Coll Biomed Engn & Instrument Sci Hangzhou 310027 Peoples R China Alibaba Cloud Hangzhou 310030 Peoples R China Zhejiang Prov Key Lab Network Multimedia Technol Hangzhou 310027 Peoples R China Minist Educ China Embedded Syst Engn Res Ctr Hangzhou 310027 Peoples R China

Enlarging input images is a straightforward and effective approach to promote small object detection. However, simple image enlargement is significantly expensive on both computations and GPU memory. In fact, small objects are usually sparsely distributed and locally clustered. Therefore, massive feature extraction computations are wasted on the non-target background area of images. Recent works have tried to pick out target-containing regions using an extra network and perform conventional object detection, but the newly introduced computation limits their final performance. In this paper, we propose to reuse the detector's backbone to conduct feature-level object-seeking and patch-slicing, which can avoid redundant feature extraction and reduce the computation cost. Incorporating with a sparse detection head, we are able to detect small objects on high-resolution inputs (e.g., 1080P or larger) for superior performance. The resulting Efficient Small object detection (ESOD) approach is a generic framework, which can be applied to both CNN- and ViT-based detectors to save the computation and GPU memory costs. Extensive experiments demonstrate the efficacy and efficiency of our method. In particular, our method consistently surpasses the SOTA detectors by a large margin (e.g., 8% gains on AP) on the representative VisDrone, UAVDT, and TinyPerson datasets. Code will be made public soon.

关键词： Feature extraction object detection Detectors Costs Head Neck Image resolution Graphics processing units Visualization Feeds Small object detection high-resolution detection filter-then-detect sparse detection

来源：评论

学校读者我要写书评

暂无评论

Large Model-Assisted Federated Learning for object detection of Autonomous Vehicles in Edge

引用

IEEE TRANSACTIONS ON VEHICULAR TECHNOLOGY 2025年第2期74卷 1839-1848页

作者： Behera, Saswat Adhikari, Mainak Menon, Varun G. Khan, Mohammad Ayoub Indian Inst Informat Technol Lucknow Lucknow 226002 India Indian Inst Sci Educ & Res Thiruvananthapuram Thiruvananthapuram 695551 India SCMS Sch Engn & Technol Karukutty 683576 India Univ Bisha Coll Comp & Informat Technol Bisha 67714 Saudi Arabia

The advancement of Autonomous Vehicles (AVs) significantly relies on the integration of Internet-of-Things technology for real-time data processing and decision-making. object detection, a critical component of AVs, necessitates utilizing machine learning techniques. However, this poses challenges including data privacy and the inefficiency of transmitting large volumes of data to a central server. To overcome these challenges, we employ Federated Learning (FL), which allows local devices to train a global model without sharing their raw data. While effective, synchronous federated learning can be time-consuming due to stragglers and often encounter device dropout issues. In contrast, asynchronous federated learning provides faster updates but frequently yields sub-optimal models compared to its synchronous counterpart. Motivated by that, in this paper, we introduce a novel FL framework that combines the strengths of both synchronous and asynchronous methods. Further, by organizing devices into a hierarchical structure, we aim to optimize model convergence while mitigating straggler and dropout problems. Moreover, deploying large models on local edge devices is impractical due to limited computational capabilities, necessitating the adoption of lightweight models to mitigate prolonged training and prediction duration. To evaluate the performance of the proposed method, we conduct a series of experiments using publicly available datasets and subsequently compare the results. Our findings demonstrate that the proposed method significantly enhances the model convergence and performance in object detection for AVs.

关键词： Computational modeling Servers Training Data models Image edge detection Federated learning Autonomous vehicles object detection autonomous vehicles (AVs) federated learning (FL) large model edge computing

来源：评论

学校读者我要写书评

暂无评论

Automatic Segmentation Annotation of Space Target Using Segment Anything Model and object detection Prompts

引用

IEEE TRANSACTIONS ON AEROSPACE AND ELECTRONIC SYSTEMS 2025年第2期61卷 4988-5001页

作者： Zhang, Zhihao Dang, Zhaohui Northwestern Polytech Univ Dept Sch Astronaut Xian 710072 Peoples R China

The efficacy of intelligent recognition and inference in space target imagery is profoundly dependent on the dataset's scale and its quality. Given the rarity of such images, labor-intensive manual annotation has been the traditional recourse. To address the demand for automatic segmentation annotation in space target imagery, we propose a groundbreaking framework utilizing the segment anything model, which leverages object detection prompts to generate precise instance boundaries. Our framework incorporates a scalable set of prompts for foreground and background classes, maximizing the zero-shot potential of our visual model. In addition, we address frequent issues with anomalous mask regions in annotations by implementing an easily integrated mask quality enhancement strategy, which has led to a precision increase of 1.0%-2.0% and an 86.7% improvement in quality. Through extensive evaluation across various public space target image datasets, our method has proven highly accurate and generalizable. To support further research, we are releasing both our specialized dataset and the project code, aiming to mitigate the limitations imposed by the scarcity of annotated training data in current on-orbit machine vision systems.

关键词： Annotations Image segmentation object segmentation Space vehicles object detection Accuracy Semantics Visualization Aerospace and electronic systems Target recognition object detection prompts segment anything model (SAM) segmentation annotation space target

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：