arXiv subjects

Li Tao

Publications and source records attributed to Li Tao.

28 records · Page 2Linked to original sources

arXiv2021

Pretext-Contrastive Learning: Toward Good Practices in Self-supervised Video Representation Leaning

Recently, pretext-task based methods are proposed one after another in self-supervised video feature learning. Meanwhile, contrastive learning methods also yield good performance. Usually, new methods can beat previous ones as claimed that they could capture "better" temporal information. However, there exist setting differences among them and it is hard to conclude which is better. It would be much more convincing in comparison if these methods have reached as closer to their performance limits as possible. In this paper, we start from one pretext-task baseline, exploring how far it can go by combining it with contrastive learning, data pre-processing, and data augmentation. A proper setting has been found from extensive experiments, with which huge improvements over the baselines can be achieved, indicating a joint optimization framework can boost both pretext task and contrastive learning. We denote the joint optimization framework as Pretext-Contrastive Learning (PCL). The other two pretext task baselines are used to validate the effectiveness of PCL. And we can easily outperform current state-of-the-art methods in the same training manner, showing the effectiveness and the generality of our proposal. It is convenient to treat PCL as a standard training strategy and apply it to many other works in self-supervised video feature learning.

cs.CV↗

arXiv2020

Self-supervised Video Representation Learning Using Inter-intra Contrastive Framework

We propose a self-supervised method to learn feature representations from videos. A standard approach in traditional self-supervised methods uses positive-negative data pairs to train with contrastive learning strategy. In such a case, different modalities of the same video are treated as positives and video clips from a different video are treated as negatives. Because the spatio-temporal information is important for video representation, we extend the negative samples by introducing intra-negative samples, which are transformed from the same anchor video by breaking temporal relations in video clips. With the proposed Inter-Intra Contrastive (IIC) framework, we can train spatio-temporal convolutional networks to learn video representations. There are many flexible options in our IIC framework and we conduct experiments by using several different configurations. Evaluations are conducted on video retrieval and video recognition tasks using the learned video representation. Our proposed IIC outperforms current state-of-the-art results by a large margin, such as 16.7% and 9.5% points improvements in top-1 accuracy on UCF101 and HMDB51 datasets for video retrieval, respectively. For video recognition, improvements can also be obtained on these two benchmark datasets. Code is available at https://github.com/BestJuly/Inter-intra-video-contrastive-learning.

cs.CV↗

arXiv2020

Motion Representation Using Residual Frames with 3D CNN

Recently, 3D convolutional networks (3D ConvNets) yield good performance in action recognition. However, optical flow stream is still needed to ensure better performance, the cost of which is very high. In this paper, we propose a fast but effective way to extract motion features from videos utilizing residual frames as the input data in 3D ConvNets. By replacing traditional stacked RGB frames with residual ones, 35.6% and 26.6% points improvements over top-1 accuracy can be obtained on the UCF101 and HMDB51 datasets when ResNet-18 models are trained from scratch. And we achieved the state-of-the-art results in this training mode. Analysis shows that better motion features can be extracted using residual frames compared to RGB counterpart. By combining with a simple appearance path, our proposal can be even better than some methods using optical flow streams.

cs.CV↗

arXiv2020

Weakly Supervised Video Summarization by Hierarchical Reinforcement Learning

Conventional video summarization approaches based on reinforcement learning have the problem that the reward can only be received after the whole summary is generated. Such kind of reward is sparse and it makes reinforcement learning hard to converge. Another problem is that labelling each frame is tedious and costly, which usually prohibits the construction of large-scale datasets. To solve these problems, we propose a weakly supervised hierarchical reinforcement learning framework, which decomposes the whole task into several subtasks to enhance the summarization quality. This framework consists of a manager network and a worker network. For each subtask, the manager is trained to set a subgoal only by a task-level binary label, which requires much fewer labels than conventional approaches. With the guide of the subgoal, the worker predicts the importance scores for video frames in the subtask by policy gradient according to both global reward and innovative defined sub-rewards to overcome the sparse problem. Experiments on two benchmark datasets show that our proposal has achieved the best performance, even better than supervised approaches.

cs.CV↗

arXiv2020

Rethinking Motion Representation: Residual Frames with 3D ConvNets for Better Action Recognition

Recently, 3D convolutional networks yield good performance in action recognition. However, optical flow stream is still needed to ensure better performance, the cost of which is very high. In this paper, we propose a fast but effective way to extract motion features from videos utilizing residual frames as the input data in 3D ConvNets. By replacing traditional stacked RGB frames with residual ones, 20.5% and 12.5% points improvements over top-1 accuracy can be achieved on the UCF101 and HMDB51 datasets when trained from scratch. Because residual frames contain little information of object appearance, we further use a 2D convolutional network to extract appearance features and combine them with the results from residual frames to form a two-path solution. In three benchmark datasets, our two-path solution achieved better or comparable performances than those using additional optical flow methods, especially outperformed the state-of-the-art models on Mini-kinetics dataset. Further analysis indicates that better motion features can be extracted using residual frames with 3D ConvNets, and our residual-frame-input path is a good supplement for existing RGB-frame-input models.

cs.CV↗

arXiv2017

Towards Universal Non-Volatile Resistance Switching in Non-metallic Monolayer Atomic Sheets

Here, we report the intriguing observation of stable non-volatile resistance switching (NVRS) in single-layer atomic sheets sandwiched between metal electrodes. NVRS is observed in the prototypical semiconducting (MX2, M=Mo, W; and X=S, Se) transitional metal dichalcogenides (TMDs), and insulating hexagonal boron nitride (h-BN), which alludes to the universality of this phenomenon in non-metallic 2D monolayers, and features forming-free switching. This observation of NVRS phenomenon, widely attributed to ionic diffusion, filament and interfacial redox in bulk oxides and electrolytes, inspires new studies on defects, ion transport and energetics at the sharp interfaces between atomically-thin sheets and conducting electrodes. From a contemporary perspective, switching is all the more unexpected in monolayers since leakage current is a fundamental limit in ultra-thin oxides. Emerging device concepts in non-volatile flexible memory fabrics, and brain-inspired (neuromorphic) computing could benefit substantially from the pervasive NVRS effect and the associated wide materials and engineering co-design opportunity. Experimentally, a 50 GHz radio-frequency (RF) monolayer switch is demonstrated, which opens up a new application for electronic zero-static power RF switching technology.

cond-mat.mes-hall↗

arXiv2017

Hybrid graphene tunneling photoconductor with interface engineering towards fast photoresponse and high responsivity

Hybrid graphene photoconductor/phototransistor has achieved giant photoresponsivity, but its response speed dramatically degrades as the expense due to the long lifetime of trapped interfacial carriers. In this work, by intercalating a large-area atomically thin MoS2 film into a hybrid graphene photoconductor, we have developed a prototype tunneling photoconductor, which exhibits a record-fast response (rising time ~17 ns) and a high responsivity (~$3\times10^4$ A/W at 635 nm and 16.8 nW illumination) across the broad spectral range. We demonstrate that the photo-excited carriers generated in silicon are transferred into graphene through a tunneling process rather than carrier drift. The atomically thin MoS2 film not only serves as tunneling layer but also passivates surface states, which in combination delivers a superior response speed (~3 order of magnitude improved than a device without MoS2 layer), while the responsivity remains high. This intriguing tunneling photoconductor integrates both fast response and high responsivity and thus has significant potential in practical applications of optoelectronic devices.

cond-mat.mes-hall↗

arXiv2015

Mott-Kondo Insulator Behavior in the Iron Oxychalcogenides

We perform a combined experimental-theoretical study of the Fe-oxychalcogenides (FeO$\emph{Ch}$) series La$_{2}$O$_{2}$Fe$_{2}$O\emph{M}$_{2}$ (\emph{M}=S, Se), which is the latest among the Fe-based materials with the potential \ to show unconventional high-T$_{c}$ superconductivity (HTSC). A combination of incoherent Hubbard features in X-ray absorption (XAS) and resonant inelastic X-ray scattering (RIXS) spectra, as well as resitivity data, reveal that the parent FeO$\emph{Ch}$ are correlation-driven insulators. To uncover microscopics underlying these findings, we perform local density approximation-plus-dynamical mean field theory (LDA+DMFT) calculations that unravel a Mott-Kondo insulating state. Based upon good agreement between theory and a range of data, we propose that FeO$\emph{Ch}$ may constitute a new, ideal testing ground to explore HTSC arising from a strange metal proximate to a novel selective-Mott quantum criticality.

cond-mat.str-el↗

arXiv2014

Toward Air-Stable Multilayer Phosphorene Thin-Films and Transistors

Few-layer black phosphorus (BP), also known as phosphorene, is poised to be the most attractive graphene analogue owing to its high mobility approaching that of graphene, and its thickness- tunable band gap that can be as large as that of molybdenum disulfide. In essence, phosphorene represents the much sought after high-mobility, large direct band gap two-dimensional layered crystal that is ideal for optoelectronics and flexible devices. However, its instability in air is of paramount concern for practical applications. Here, we demonstrate air-stable BP devices with dielectric and hydrophobic encapsulation. Microscopy, spectroscopy, and transport techniques were employed to elucidate the aging mechanism, which can initiate from the BP surface for bare samples, or edges for samples with thin dielectric coating highlighting the ineffectiveness of conventional scaled dielectrics. Our pioneering months-long studies indicate that a double layer of Al2O3 and hydrophobic fluoropolymer affords BP devices and transistors with indefinite air-stability for the first time, overcoming a critical material challenge for applied research and development.

cond-mat.mtrl-sci↗

arXiv2012

Uniform wafer-scale synthesis of graphene on evaporated Cu (111) film with quality comparable to exfoliated monolayer

Monolayer graphene has been grown on crystallized Cu (111) films on standard oxidized Si 100 mm wafers. The monolayer graphene demonstrates high uniformity (>97% coverage), with immeasurable defects (>95% defect-negligible) across the entire wafer. Key to these results is the phase transition of evaporated copper films from amorphous to crystalline at the growth temperature as corroborated by X-ray diffraction and electron backscatter diffraction. Noticeably, phase transition of copper film is observed on technologically ubiquitous oxidized Si wafer where the oxide is a standard amorphous thermal oxide. Ion mass spectroscopy indicates that the copper films can be purposely hydrogen-enriched during a hydrogen anneal which subsequently affords graphene growth with a sole carbonaceous precursor for low defect densities. Owing to the strong hexagonal lattice match, the graphene domains align to the Cu (111) domains, suggesting a pathway for increasing the graphene grains by maximizing the copper grain sizes. Fabricated graphene transistors on a flexible polyimide film yield a peak carrier mobility ~4,930 cm2/Vs.

cond-mat.mtrl-sci↗