相关参考文献
注意:仅列出部分参考文献,下载原文获取全部文献信息。
Article
Computer Science, Artificial Intelligence
Hong Yin et al.
Summary: This paper presents a Dual Context Network (DCNet) to address the challenge of real-time semantic segmentation, considering both segmentation accuracy and inference speed. DCNet consists of two independent sub-networks: a Region Context Network and a Pixel Context Network. The Region Context Network has a low-resolution input and a features re-weighting module to achieve sufficient receptive field. The Pixel Context Network, on the other hand, captures the location dependencies of each pixel through a location attention module to assist the main network in recovering spatial detail. A contextual feature fusion is introduced to combine the output features of these two sub-networks. Experimental results show that DCNet can achieve high-quality segmentation while maintaining a high speed. Specifically, using ResNet50 as the backbone, it achieves a Mean IOU of 76.1% at a speed of 82 FPS on a single GTX 2080Ti GPU, and a Mean IOU of 71.2% at a speed of 142 FPS using ResNet18 as the backbone.
MACHINE VISION AND APPLICATIONS
(2023)
Article
Computer Science, Information Systems
Twinkle Tiwari et al.
Summary: This paper presents a modified-UNet model for segmenting vehicles in images of road traffic with congested and unstructured traffic patterns. By using convolutions, inception modules, and batch normalization in the encoding and decoding steps, the performance of the model is improved. Experimental results on two datasets demonstrate that the proposed model outperforms other models in terms of intersection over union.
MULTIMEDIA TOOLS AND APPLICATIONS
(2023)
Article
Computer Science, Artificial Intelligence
Tsung-Han Tsai et al.
Article
Computer Science, Artificial Intelligence
Peng Ding et al.
Summary: Semantic segmentation helps improve the understanding of complex scenes and assists unmanned systems in perceiving scene content. To address the issues of information loss and edge blur in semantic segmentation for complex scenes, we propose a modified version of Deeplabv3+ with improved ASPP and fusion module. Our proposed modules significantly enhance the accuracy of semantic segmentation and achieve comparable results with state-of-the-art algorithms.
JOURNAL OF REAL-TIME IMAGE PROCESSING
(2023)
Article
Engineering, Civil
Guoan Xu et al.
Summary: This paper introduces a lightweight real-time semantic segmentation network called LETNet, which combines U-shaped CNN with Transformer effectively to compensate for respective deficiencies. The elaborately designed Lightweight Dilated Bottleneck (LDB) module and Feature Enhancement (FE) module simultaneously have a positive impact on training from scratch. Extensive experiments on challenging datasets demonstrate that LETNet achieves superior performances in accuracy and efficiency balance.
IEEE TRANSACTIONS ON INTELLIGENT TRANSPORTATION SYSTEMS
(2023)
Article
Geochemistry & Geophysics
Ruigang Niu et al.
Summary: This research proposes a novel attention-based framework named hybrid multiple attention network (HMANet) for semantic segmentation in remote sensing images. The HMANet adaptively captures global correlations by introducing class augmented attention and region shuffle attention modules, improving the efficiency and effectiveness of the self-attention mechanism.
IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING
(2022)
Review
Computer Science, Artificial Intelligence
Xiuling Zhang et al.
Summary: In this paper, a lightweight attention-guided asymmetric network (LAANet) is proposed to achieve a good trade-off between segmentation accuracy, inference speed, and model size in real-world scenarios such as robot navigation and autonomous driving.
NEURAL COMPUTING & APPLICATIONS
(2022)
Article
Computer Science, Software Engineering
Zhiming Cheng et al.
Summary: A contour-aware semantic segmentation network, an extension of Unet, was proposed in this paper for medical image segmentation. The method includes a semantic branch and a detail branch, utilizing MulBlock module and CAM module to enhance network performance.
Article
Computer Science, Software Engineering
Kang Wang et al.
Summary: This paper introduces a novel semantic segmentation method based on an asymmetric encoder-decoder network structure, which combines depth-wise separable asymmetric convolution and attention mechanism to significantly reduce computation cost while maintaining accuracy, achieving a balance between segmentation performance and speed.
Article
Computer Science, Software Engineering
Min Jiang et al.
Summary: The paper proposes a Sparse Attention Model combined with a powerful multi-task feature extraction network to reduce computing resource consumption in semantic segmentation. By using a Class Attention Module, the model ensures that query vectors capture dense contextual information efficiently.
Article
Computer Science, Information Systems
Yanping Tang et al.
Summary: An efficient semantic segmentation method based on attention gate and multi-layer fusion is proposed in this paper, which enhances the emphasis on foreground features and the integration of semantic information, achieving a mean IOU of 72.9% on the CamVid test dataset at a speed of 43 FPS.
MULTIMEDIA TOOLS AND APPLICATIONS
(2022)
Article
Computer Science, Artificial Intelligence
Zechao Li et al.
Summary: This study proposes a novel Context-based Tandem Network (CTNet) that explores spatial and channel contextual information for semantic segmentation. The CTNet demonstrates superior performance by adaptively integrating the results of two context modules, leading to improved learning representations.
IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE
(2022)
Article
Engineering, Electrical & Electronic
Xi Weng et al.
Summary: This paper presents a Stage-aware Feature Alignment Network (SFANet) based on an encoder-decoder structure for real-time semantic segmentation of street scenes. By introducing a stage-aware Feature Enhancement Block (FEB) and an auxiliary training strategy, SFANet achieves a good balance between accuracy and speed.
IEEE TRANSACTIONS ON CIRCUITS AND SYSTEMS FOR VIDEO TECHNOLOGY
(2022)
Article
Environmental Sciences
Runrui Liu et al.
Summary: In this paper, an improved deep learning model RAANet is proposed, which constructs a new residual ASPP by embedding attention module and residual structure into ASPP for multi-scale semantic information and improved classification accuracy of land use in remote sensing images.
Article
Automation & Control Systems
Wanxia Deng et al.
Summary: Unsupervised domain adaptation (UDA) involves learning a classifier for an unlabeled target domain by transferring knowledge from a labeled source domain with a related but different distribution. The deep ladder-suppression network (DLSN) proposed in this study is designed to better learn cross-domain shared content by suppressing domain-specific variations. Experimental results demonstrate that the DLSN consistently and significantly improves the performance of various popular UDA frameworks.
IEEE TRANSACTIONS ON CYBERNETICS
(2022)
Proceedings Paper
Computer Science, Artificial Intelligence
Tianfei Zhou et al.
Summary: This study proposes a nonparametric semantic segmentation solution based on non-learnable prototypes. By optimizing the arrangement between embedded pixels and prototypes, dense prediction is achieved. Compared to prior parametric methods, our approach does not require learning individual weights/query vectors for each class, but represents each class using non-learnable prototypes. The nonparametric framework performs well on multiple datasets and model architectures.
2022 IEEE/CVF CONFERENCE ON COMPUTER VISION AND PATTERN RECOGNITION (CVPR 2022)
(2022)
Article
Computer Science, Information Systems
Wanxia Deng et al.
Summary: Unsupervised Domain Adaptation (UDA) aims to transfer knowledge from a labeled source domain to an unlabeled target domain with a related but different distribution. Existing methods have made progress in aligning the two domains in latent feature space, but they mainly focus on adapting the entire image and ignore the negative effect of uninformative domain-specific variations on learned features. To address this issue, this paper proposes a novel component called Informative Feature Disentanglement (IFD) equipped with adversarial networks or metric discrepancy models. The proposed IFDAN and IFDMN models refine informative features before adaptation, effectively disentangling them from uninformative domain-specific variations. Extensive experimental results on three gold-standard domain adaptation datasets demonstrate the effectiveness of the proposed IFDAN and IFDMN models for UDA.
IEEE TRANSACTIONS ON MULTIMEDIA
(2022)
Article
Computer Science, Artificial Intelligence
Shijie Hao et al.
Summary: The proposed SGCPNet utilizes spatial-detail guided context propagation to achieve real-time semantic segmentation by effectively reconstructing lost spatial information using spatial details from shallow layers, thus improving model efficiency and maintaining segmentation accuracy.
IEEE TRANSACTIONS ON NEURAL NETWORKS AND LEARNING SYSTEMS
(2022)
Article
Computer Science, Artificial Intelligence
Jingdong Wang et al.
Summary: The High-Resolution Network (HRNet) maintains high-resolution representations and exchanges information across resolutions, resulting in superior performance in various applications such as human pose estimation, semantic segmentation, and object detection.
IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE
(2021)
Article
Computer Science, Information Systems
Ashish Sinha et al.
Summary: This paper introduces a new architecture that improves the accuracy and reliability of medical image segmentation by capturing richer contextual dependencies through guided self-attention mechanisms. Compared to other state-of-the-art models, our model demonstrates better segmentation performance, showcasing the effectiveness of our approach.
IEEE JOURNAL OF BIOMEDICAL AND HEALTH INFORMATICS
(2021)
Article
Computer Science, Artificial Intelligence
Peng Sun et al.
Summary: This paper introduces a joint search framework, AutoRTNet, which automates key properties in semantic segmentation and achieves the best trade-off between accuracy and speed on Cityscapes and CamVid datasets.
INTERNATIONAL JOURNAL OF COMPUTER VISION
(2021)
Article
Computer Science, Artificial Intelligence
Mingxi Zhuang et al.
Summary: This paper proposes a lightweight network LRDNet for a better balance between computational speed and segmentation accuracy. The network includes a refined dual attention decoder, asymmetric module, and deep convolution to improve efficiency in feature extraction, achieving impressive results on the Cityscapes dataset.
Article
Engineering, Civil
Genshun Dong et al.
Summary: The proposed real-time high-performance DCNN-based method achieves a good trade-off between accuracy and speed for semantic segmentation of urban street scenes. By exploiting different scales of pooling operations, combining shallow convolutional layers, and utilizing feature fusion networks, the method effectively achieves image segmentation with excellent performance at real-time speed.
IEEE TRANSACTIONS ON INTELLIGENT TRANSPORTATION SYSTEMS
(2021)
Article
Computer Science, Artificial Intelligence
Changqian Yu et al.
Summary: Separating low-level details and high-level semantics is key to achieving high accuracy and efficiency in real-time semantic segmentation. The proposed architecture, called Bilateral Segmentation Network (BiSeNet V2), effectively handles feature representations through detail and semantics branches, striking a balance between speed and accuracy to outperform existing methods.
INTERNATIONAL JOURNAL OF COMPUTER VISION
(2021)
Article
Robotics
Ping Hu et al.
Summary: This paper introduces a novel deep CNN architecture for semantic segmentation of high-resolution images and videos, achieving state-of-the-art performance with the use of fast spatial attention and additional spatial reduction. Experimental results demonstrate superior accuracy and speed compared to existing approaches.
IEEE ROBOTICS AND AUTOMATION LETTERS
(2021)
Proceedings Paper
Computer Science, Hardware & Architecture
Cheng Liu et al.
2020 INTERNATIONAL SYMPOSIUM ON AUTONOMOUS SYSTEMS (ISAS)
(2020)
Article
Computer Science, Artificial Intelligence
Liang-Chieh Chen et al.
IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE
(2018)
Proceedings Paper
Computer Science, Artificial Intelligence
Francois Chollet
30TH IEEE CONFERENCE ON COMPUTER VISION AND PATTERN RECOGNITION (CVPR 2017)
(2017)
Article
Computer Science, Artificial Intelligence
Olga Russakovsky et al.
INTERNATIONAL JOURNAL OF COMPUTER VISION
(2015)