Representation learning from unlabeled skeleton data is a challenging task.Prior unsupervised learning algorithms mainly rely on the modeling ability of recurrent neural networks to extract the action representations....Representation learning from unlabeled skeleton data is a challenging task.Prior unsupervised learning algorithms mainly rely on the modeling ability of recurrent neural networks to extract the action representations.However,the structural information of the skeleton data,which also plays a critical role in action recognition,is rarely explored in existing unsupervised methods.To deal with this limitation,we propose a novel twostream autoencoder network to combine the topological information with temporal information of skeleton data.Specifically,we encode the graph structure by graph convolutional network(GCN)and integrate the extracted GCN-based representations into the gate recurrent unit stream.Then we design a transfer module to merge the representations of the two streams adaptively.According to the characteristics of the two-stream autoencoder,a unified loss function composed of multiple tasks is proposed to update the learnable parameters of our model.Comprehensive experiments on NW-UCLA,UWA3D,and NTU-RGBD 60 datasets demonstrate that our proposed method can achieve an excellent performance among the unsupervised skeleton-based methods and even perform a similar or superior performance over numerous supervised skeleton-based methods.展开更多
This paper proposes a novel method for early action prediction based on 3D skeleton data. Our method combines the advantages of graph convolutional networks (GCNs) and adversarial learning to avoid the problems of ins...This paper proposes a novel method for early action prediction based on 3D skeleton data. Our method combines the advantages of graph convolutional networks (GCNs) and adversarial learning to avoid the problems of insufficient spatio-temporal feature extraction and difficulty in predicting actions in the early execution stage of actions. In our method, GCNs, which have outstanding performance in the field of action recognition, are used to extract the spatio-temporal features of the skeleton. The model learns how to optimize the feature distribution of partial videos from the features of full videos through adversarial learning. Experiments on two challenging action prediction datasets show that our method performs well on skeleton-based early action prediction. State-of-the-art performance is reported in some observation ratios.展开更多
Chiral luminescence materials have potential applications in the field of three-dimensional displays due to their circularly polarized luminescence(CPL)characteristics.However,the further development of circularly pol...Chiral luminescence materials have potential applications in the field of three-dimensional displays due to their circularly polarized luminescence(CPL)characteristics.However,the further development of circularly polarized organic light-emitting diodes(CP-OLEDs)needs to meet the requirements of high efficiency,high color purity,low cost,and high dissymmetry factor(gPLor gEL),chiral multiple resonance thermally activated delayed fluorescence(MR-TADF)materials are considered as candidates in these aspects.Herein,based on a pair of chiral spirofluorene precursors,two pairs of high-performance chiral MR-TADF emitters((R/S)-p-Spiro-DtBuCzB and(R/S)-m-Spiro-DtBuCzB)are developed,which exhibit strong emissions peaking at 491 and 502 nm in toluene with full-width at half-maximum values of 25 and 33 nm,respectively.In addition,small singlet–triplet energy gaps of 0.15 and 0.10 eV with high absolute photoluminescence efficiencies of 95.0%and 96.7%are observed for p-Spiro-DtBuCzB and m-Spiro-DtBuCzB molecules,respectively.OLEDs based on p-Spiro-DtBuCzB and m-Spiro-DtBuCzB display high maximum external quantum efficiencies of 29.6%and 33.8%,respectively.Most importantly,CP-OLEDs present symmetric circularly polarized electroluminescence spectra with|gEL|factors of 3.36×10^(-4)and 7.66×10^(-4)for devices based on(R/S)-p-Spiro-DtBuCzB and(R/S)-m-Spiro-DtBuCzB enantiomers,respectively.展开更多
This paper presents a rigidity-preserving morphing technique that blends a pair of 2D shapes in a controllable manner. The morphing is controllable in two aspects: 1) motion dynamics in the interpolation sequences can...This paper presents a rigidity-preserving morphing technique that blends a pair of 2D shapes in a controllable manner. The morphing is controllable in two aspects: 1) motion dynamics in the interpolation sequences can be effectively enhanced through an intuitive skeleton control and 2) not only the boundaries but also the interior features of the source and target shapes are precisely aligned during the morphing. We introduce a new compatible triangulation algorithm to decompose the source and target shapes into isomorphic triangles. Moreover, a robust and motion-controllable rigiditypreserving transformation scheme is proposed to blend the compatible triangulations, ultimately leading to a morphing sequence which is appearance-preserving and with the desired motion dynamics. Our approach constitutes a powerful and easy-to-use morphing tool for two-dimensional animation. We demonstrate its versatility, effectiveness and visual accuracy through a variety of examples and comparisons to prior work.展开更多
文摘Representation learning from unlabeled skeleton data is a challenging task.Prior unsupervised learning algorithms mainly rely on the modeling ability of recurrent neural networks to extract the action representations.However,the structural information of the skeleton data,which also plays a critical role in action recognition,is rarely explored in existing unsupervised methods.To deal with this limitation,we propose a novel twostream autoencoder network to combine the topological information with temporal information of skeleton data.Specifically,we encode the graph structure by graph convolutional network(GCN)and integrate the extracted GCN-based representations into the gate recurrent unit stream.Then we design a transfer module to merge the representations of the two streams adaptively.According to the characteristics of the two-stream autoencoder,a unified loss function composed of multiple tasks is proposed to update the learnable parameters of our model.Comprehensive experiments on NW-UCLA,UWA3D,and NTU-RGBD 60 datasets demonstrate that our proposed method can achieve an excellent performance among the unsupervised skeleton-based methods and even perform a similar or superior performance over numerous supervised skeleton-based methods.
基金supported by the Natural Science Foundation of Xinjiang Uygur Autonomous Region of China under Grant No.2022D01A59Xinjiang Uygur Autonomous Region University Scientific Research Project(Key Natural Science Project)under Grant No.XJEDU2021I029+1 种基金the National Natural Science Foundation of China under Grant No.U20A20167the Project of Hebei Key Laboratory of Software Engineering under Grant No.22567637H.
文摘This paper proposes a novel method for early action prediction based on 3D skeleton data. Our method combines the advantages of graph convolutional networks (GCNs) and adversarial learning to avoid the problems of insufficient spatio-temporal feature extraction and difficulty in predicting actions in the early execution stage of actions. In our method, GCNs, which have outstanding performance in the field of action recognition, are used to extract the spatio-temporal features of the skeleton. The model learns how to optimize the feature distribution of partial videos from the features of full videos through adversarial learning. Experiments on two challenging action prediction datasets show that our method performs well on skeleton-based early action prediction. State-of-the-art performance is reported in some observation ratios.
基金supported by the National Natural Science Foundation of China(92256304,U23A20593)。
文摘Chiral luminescence materials have potential applications in the field of three-dimensional displays due to their circularly polarized luminescence(CPL)characteristics.However,the further development of circularly polarized organic light-emitting diodes(CP-OLEDs)needs to meet the requirements of high efficiency,high color purity,low cost,and high dissymmetry factor(gPLor gEL),chiral multiple resonance thermally activated delayed fluorescence(MR-TADF)materials are considered as candidates in these aspects.Herein,based on a pair of chiral spirofluorene precursors,two pairs of high-performance chiral MR-TADF emitters((R/S)-p-Spiro-DtBuCzB and(R/S)-m-Spiro-DtBuCzB)are developed,which exhibit strong emissions peaking at 491 and 502 nm in toluene with full-width at half-maximum values of 25 and 33 nm,respectively.In addition,small singlet–triplet energy gaps of 0.15 and 0.10 eV with high absolute photoluminescence efficiencies of 95.0%and 96.7%are observed for p-Spiro-DtBuCzB and m-Spiro-DtBuCzB molecules,respectively.OLEDs based on p-Spiro-DtBuCzB and m-Spiro-DtBuCzB display high maximum external quantum efficiencies of 29.6%and 33.8%,respectively.Most importantly,CP-OLEDs present symmetric circularly polarized electroluminescence spectra with|gEL|factors of 3.36×10^(-4)and 7.66×10^(-4)for devices based on(R/S)-p-Spiro-DtBuCzB and(R/S)-m-Spiro-DtBuCzB enantiomers,respectively.
基金supported by the National Natural Science Foundation of China under Grant Nos.61003189,U1609215 and 61472363the US National Science Foundation under Grant Nos. 0915933,0937586,and 1647200.
文摘This paper presents a rigidity-preserving morphing technique that blends a pair of 2D shapes in a controllable manner. The morphing is controllable in two aspects: 1) motion dynamics in the interpolation sequences can be effectively enhanced through an intuitive skeleton control and 2) not only the boundaries but also the interior features of the source and target shapes are precisely aligned during the morphing. We introduce a new compatible triangulation algorithm to decompose the source and target shapes into isomorphic triangles. Moreover, a robust and motion-controllable rigiditypreserving transformation scheme is proposed to blend the compatible triangulations, ultimately leading to a morphing sequence which is appearance-preserving and with the desired motion dynamics. Our approach constitutes a powerful and easy-to-use morphing tool for two-dimensional animation. We demonstrate its versatility, effectiveness and visual accuracy through a variety of examples and comparisons to prior work.
