End-to-end object detection methods have attracted extensive interest recently since they alleviate the need for complicated human-designed components and simplify the detection pipeline.However,these methods suffer f...End-to-end object detection methods have attracted extensive interest recently since they alleviate the need for complicated human-designed components and simplify the detection pipeline.However,these methods suffer from slower training convergence and inferior detection performance compared to conventional detectors,as their feature fusion and selection processes are constrained by insufficient positive supervision.To address this issue,we introduce a novel query-selection encoder(QsE)designed for end-to-end object detectors to improve the training convergence speed and detection accuracy.QsE is composed of multiple encoder layers stacked on top of the backbone.A lightweight head network is added after each encoder layer to continuously optimize features in a cascading manner,providing more positive supervision for efficient training.Additionally,a hierarchical feature-aware attention(HFA)mechanism is incorporated in each encoder layer,including in-and cross-level feature attention,to enhance the interaction between features from different levels.HFA can effectively suppress similar feature representations and highlight discriminative ones,thereby accelerating the feature selection process.Our method is highly versatile in accommodating both CNN-and Transformer-based detectors.Extensive experiments were conducted on the popular benchmark datasets MS COCO,CrowdHuman,and PASCAL VOC to demonstrate the effectiveness of our method.The results showed that CNN-and Transformer-based detectors using QSE can achieve better end-to-end performance within fewer training epochs.展开更多
In the area of 3D digital engineering and 3D digital geometry processing, shape simplification is an important task to reduce their requirement of large memory and high time complexity. By incorporating the content-aw...In the area of 3D digital engineering and 3D digital geometry processing, shape simplification is an important task to reduce their requirement of large memory and high time complexity. By incorporating the content-aware visual salience measure of a polygonal mesh into simplification operation, a novel feature-aware shape simplification approach is presented in this paper. Owing to the robust extraction of relief heights on 3D highly detailed meshes, our visual salience measure is defined by a center-surround operator on Gaussian-weighted relief heights in a scale-dependent manner. Guided by our visual salience map, the feature-aware shape simplification algorithm can be performed by weighting the high-dimensional feature space quadric error metric of vertex pair contractions with the weight map derived from our visual salience map. The weighted quadric error metric is calculated in a six-dimensional feature space by combining the position and normal information of mesh vertices. Experimental results demonstrate that our visual salience guided shape simplification scheme can adaptively and effectively re-sample the underlying models in a feature-aware manner, which can account for the visually salient features of the complex shapes and thus yield better visual fidelity.展开更多
Mesh segmentation is a fundamental and critical task in mesh processing,and it has been studied extensively in computer graphics and geometric modeling communities.However,current methods are not well suited for segme...Mesh segmentation is a fundamental and critical task in mesh processing,and it has been studied extensively in computer graphics and geometric modeling communities.However,current methods are not well suited for segmenting large meshes which are now common in many applications.This paper proposes a new spectral segmentation method specifically designed for large meshes inspired by multi-resolution representations.Building on edge collapse operators and progressive mesh representations,we first devise a feature-aware simplification algorithm that can generate a coarse mesh which keeps the same topology as the input mesh and preserves as many features of the input mesh as possible.Then,using the spectral segmentation method proposed in Tong et al.(IEEE Trans Vis Comput Graph 26(4):1807–1820,2020),we perform partition on the coarse mesh to obtain a coarse segmentation which mimics closely the desired segmentation of the input mesh.By reversing the simplification process through vertex split operators,we present a fast algorithm which maps the coarse segmentation to the input mesh and therefore obtain an initial segmentation of the input mesh.Finally,to smooth some jaggy boundaries between adjacent parts of the initial segmentation or align with the desired boundaries,we propose an efficient method to evolve those boundaries driven by geodesic curvature flows.As demonstrated by experimental results on a variety of large meshes,our method outperforms the state-of-the-art segmentation method in terms of not only speed but also usability.展开更多
基金Project supported by Zhejiang University Robotics Institute(No.K12106)。
文摘End-to-end object detection methods have attracted extensive interest recently since they alleviate the need for complicated human-designed components and simplify the detection pipeline.However,these methods suffer from slower training convergence and inferior detection performance compared to conventional detectors,as their feature fusion and selection processes are constrained by insufficient positive supervision.To address this issue,we introduce a novel query-selection encoder(QsE)designed for end-to-end object detectors to improve the training convergence speed and detection accuracy.QsE is composed of multiple encoder layers stacked on top of the backbone.A lightweight head network is added after each encoder layer to continuously optimize features in a cascading manner,providing more positive supervision for efficient training.Additionally,a hierarchical feature-aware attention(HFA)mechanism is incorporated in each encoder layer,including in-and cross-level feature attention,to enhance the interaction between features from different levels.HFA can effectively suppress similar feature representations and highlight discriminative ones,thereby accelerating the feature selection process.Our method is highly versatile in accommodating both CNN-and Transformer-based detectors.Extensive experiments were conducted on the popular benchmark datasets MS COCO,CrowdHuman,and PASCAL VOC to demonstrate the effectiveness of our method.The results showed that CNN-and Transformer-based detectors using QSE can achieve better end-to-end performance within fewer training epochs.
基金Project supported by the National Natural Science Foundation of China(No.61272309)the Key Laboratory of Visual Media Intelligent Process Technology of Zhejiang Province,China(No.2011E10003)
文摘In the area of 3D digital engineering and 3D digital geometry processing, shape simplification is an important task to reduce their requirement of large memory and high time complexity. By incorporating the content-aware visual salience measure of a polygonal mesh into simplification operation, a novel feature-aware shape simplification approach is presented in this paper. Owing to the robust extraction of relief heights on 3D highly detailed meshes, our visual salience measure is defined by a center-surround operator on Gaussian-weighted relief heights in a scale-dependent manner. Guided by our visual salience map, the feature-aware shape simplification algorithm can be performed by weighting the high-dimensional feature space quadric error metric of vertex pair contractions with the weight map derived from our visual salience map. The weighted quadric error metric is calculated in a six-dimensional feature space by combining the position and normal information of mesh vertices. Experimental results demonstrate that our visual salience guided shape simplification scheme can adaptively and effectively re-sample the underlying models in a feature-aware manner, which can account for the visually salient features of the complex shapes and thus yield better visual fidelity.
基金supported by the National Natural Science Foundation of China(Nos.61877056,61972368)the Anhui Provincial Natural Science Foundation,PR China(No.1908085QA11).
文摘Mesh segmentation is a fundamental and critical task in mesh processing,and it has been studied extensively in computer graphics and geometric modeling communities.However,current methods are not well suited for segmenting large meshes which are now common in many applications.This paper proposes a new spectral segmentation method specifically designed for large meshes inspired by multi-resolution representations.Building on edge collapse operators and progressive mesh representations,we first devise a feature-aware simplification algorithm that can generate a coarse mesh which keeps the same topology as the input mesh and preserves as many features of the input mesh as possible.Then,using the spectral segmentation method proposed in Tong et al.(IEEE Trans Vis Comput Graph 26(4):1807–1820,2020),we perform partition on the coarse mesh to obtain a coarse segmentation which mimics closely the desired segmentation of the input mesh.By reversing the simplification process through vertex split operators,we present a fast algorithm which maps the coarse segmentation to the input mesh and therefore obtain an initial segmentation of the input mesh.Finally,to smooth some jaggy boundaries between adjacent parts of the initial segmentation or align with the desired boundaries,we propose an efficient method to evolve those boundaries driven by geodesic curvature flows.As demonstrated by experimental results on a variety of large meshes,our method outperforms the state-of-the-art segmentation method in terms of not only speed but also usability.
