Search

Figures

Although previous co-speech gesture generation methods are able to synthesize motions in line with speech content, it is still not enough to handle diverse and complicated motion distribution. The key challenges are: 1) the one-to-many nature between the speech content and gestures; 2) the correlation modeling between the body joints. In this paper, we present a novel framework (EMoG) to tackle the above challenges with denoising diffusion models: 1) To alleviate the one-to-many problem, we incorporate emotion clues to guide the generation process, making the generation much easier; 2) To model joint correlation, we propose to decompose the difficult gesture generation into two sub-problems: joint correlation modeling and temporal dynamics modeling. Then, the two sub-problems are explicitly tackled with our proposed Joint Correlation-aware transFormer (JCFormer). Through extensive evaluations, we demonstrate that our proposed method surpasses previous state-of-the-art approaches, offering substantial superiority in gesture synthesis., Comment: under review

As a recent noticeable topic, domain generalization aims to learn a generalizable model on multiple source domains, which is expected to perform well on unseen test domains. Great efforts have been made to learn domain-invariant features by aligning distributions across domains. However, existing works are often designed based on some relaxed conditions which are generally hard to satisfy and fail to realize the desired joint distribution alignment. In this paper, we propose a novel domain generalization method, which originates from an intuitive idea that a domain-invariant classifier can be learned by minimizing the KL-divergence between posterior distributions from different domains. To enhance the generalizability of the learned classifier, we formalize the optimization objective as an expectation computed on the ground-truth marginal distribution. Nevertheless, it also presents two obvious deficiencies, one of which is the side-effect of entropy increase in KL-divergence and the other is the unavailability of ground-truth marginal distributions. For the former, we introduce a term named maximum in-domain likelihood to maintain the discrimination of the learned domain-invariant representation space. For the latter, we approximate the ground-truth marginal distribution with source domains under a reasonable convex hull assumption. Finally, a Constrained Maximum Cross-domain Likelihood (CMCL) optimization problem is deduced, by solving which the joint distributions are naturally aligned. An alternating optimization strategy is carefully designed to approximately solve this optimization problem. Extensive experiments on four standard benchmark datasets, i.e., Digits-DG, PACS, Office-Home and miniDomainNet, highlight the superior performance of our method.

Most existing image restoration networks are designed in a disposable way and catastrophically forget previously learned distortions when trained on a new distortion removal task. To alleviate this problem, we raise the novel lifelong image restoration problem for blended distortions. We first design a base fork-join model in which multiple pre-trained expert models specializing in individual distortion removal task work cooperatively and adaptively to handle blended distortions. When the input is degraded by a new distortion, inspired by adult neurogenesis in human memory system, we develop a neural growing strategy where the previously trained model can incorporate a new expert branch and continually accumulate new knowledge without interfering with learned knowledge. Experimental results show that the proposed approach can not only achieve state-of-the-art performance on blended distortions removal tasks in both PSNR/SSIM metrics, but also maintain old expertise while learning new restoration tasks., Comment: ECCV2020 accepted

As the flying wing layout unmanned aerial vehicle (uav) extensive research and task environment increasingly complex, Yu Feiyi layout unmanned aerial vehicle (uav) for fault tolerant control gradually become the main technical means of the flight control, using the established mathematical model of the flying wing uav longitudinal layout setting the actuator failure effect, is in the nature of adaptive control allocation fault-tolerant algorithm is given, and MATLAB/simulink simulation is carried out for uav longitudinal motion, realize the rapid and stable, the control command and response to complete the nonlinear fault-tolerant control of flying wing uavs.

Face restoration from low resolution and noise is important for applications of face analysis recognition. However, most existing face restoration models omit the multiple scale issues in face restoration problem, which is still not well-solved in research area. In this paper, we propose a Sequential Gating Ensemble Network (SGEN) for multi-scale noise robust face restoration issue. To endow the network with multi-scale representation ability, we first employ the principle of ensemble learning for SGEN network architecture designing. The SGEN aggregates multi-level base-encoders and base-decoders into the network, which enables the network to contain multiple scales of receptive field. Instead of combining these base-en/decoders directly with non-sequential operations, the SGEN takes base-en/decoders from different levels as sequential data. Specifically, it is visualized that SGEN learns to sequentially extract high level information from base-encoders in bottom-up manner and restore low level information from base-decoders in top-down manner. Besides, we propose to realize bottom-up and top-down information combination and selection with Sequential Gating Unit (SGU). The SGU sequentially takes information from two different levels as inputs and decides the output based on one active input. Experiment results on benchmark dataset demonstrate that our SGEN is more effective at multi-scale human face restoration with more image details and less noise than state-of-the-art image restoration models. Further utilizing adversarial training scheme, SGEN also produces more visually preferred results than other models under subjective evaluation., Comment: 11 pages, 15 figures. arXiv admin note: substantial text overlap with arXiv:1805.02164

Ru/MgO-CeO2 catalysts were prepared by co-precipitation with different precursors,including K2RuO4,Ru(Ac)3,and RuCl3,the obtained catalysts were characterized by X-ray diffraction,X-ray fluorescence,N2 physisorption,temperature-programmed reduction,and CO chemisorption. The results showed that the ammonia synthesis activity of the Ru/MgO-CeO2 catalyst prepared with K2RuO4 was 16.0% at 10 MPa,450 °C,and 10 000 h-1,which was higher than those obtained with the other two precursors. The turnover frequency (TOF) of the Ru/MgO-CeO2 catalyst prepared with K2RuO4 was higher than those obtained with the other precursors under the same reaction conditions. The influence of precursors and preparation methods on the structure and performance of the catalysts was studied. The ratio of Ru/MgO-CeO2 catalysts prepared with K2RuO4,Ru(Ac)3,and RuCl3 as precursors for ammonia synthesis turnover frequency is 1.33:1.05:1 under 10 MPa,425 °C,and 10 000 h-1.

A plasmid carrying the lambda PL promoter was constructed to express efficiently unfused human alpha D-interferon (HuIFN-alpha D) in Escherichia coli using a TGATG site in its signal sequence, which occurs also in the lambda DNA sequence. The unfused nature of HuIFN-alpha D expressed by pBV867 in E. coli (BMH 71-18) was confirmed by the following evidence: first, the purified IFN showed a single band of 19.5K in sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE); a 27K band, representing lambda N-HuIFN-alpha D fusion protein, was not detected. Second, the peak of IFN activity coincided with the 19.5K protein. Third, the peak of absorbent material for human leukocyte IFN antibody coincided with that of IFN activity. Finally, amino-terminal sequencing of purified IFN demonstrated an unfused HuIFN-alpha D. This suggests that E. coli is able to process the signal sequence of HuIFN-alpha D. Studies on the mechanism of "translational coupling" initiation of gene expression were carried out by the construction of two hybrid plasmids and titration of the IFN activities produced by them. The level of expression by the ATG-TGATG initiation mode was found to be six times higher than that of the single ATG mode.

Programmed death-1 inhibitors are approved for patients with relapsed or refractory classic Hodgkin lymphoma (RRcHL). We present the 2-year follow-up of the phase 2 KEYNOTE-087 study of pembrolizumab in 210 patients, based on HL progression after autologous stem cell transplantation (ASCT) and subsequent brentuximab vedotin (BV; cohort 1); salvage chemotherapy and BV, with ineligibility for SCT owing to chemorefractory disease (cohort 2); and progression after SCT without BV (cohort 3). With a median follow-up of 27.6 months, the objective response rate (ORR) by blinded independent central review was 71.9% (95% CI, 65.3-77.9), the complete response rate (CRR) was 27.6%, and the partial response (PR) rate was 44.3%. Median duration of response was 16.5 months (range, 0.01 to 27.01 [1, no progressive disease at last assessment]) in all patients, 22.1 months in cohort 1, 11.1 months in cohort 2, and 24.4 months in cohort 3. Median progression-free survival was not reached in all patients with CR: 13.8 months (95% CI, 12.0-22.1) for patients with PR and 10.9 months (95% CI, 5.6-11.1) for patients with stable disease. Median overall survival was not reached in all patients or in any cohort. Treatment-related adverse events (TRAEs) of any grade occurred in 153 (72.9%) patients; grades 3 and 4 occurred in 25 (12.0%) patients; none resulted in death. Results confirmed effective antitumor activity, durability of response, and manageable safety of pembrolizumab monotherapy in RRcHL, regardless of prior treatment and including chemoresistant cHL. This trial was registered at www.clinicaltrials.gov as #NCT02453594.