Search

Video question answering (VideoQA) is challenging as it requires reasoning about natural language and multimodal interactive relations. Most existing methods apply attention mechanisms to extract interactions between the question and the video or to extract effective spatio-temporal relational representations. However, these methods neglect the implication of relations between intra- and inter-modal interactions for multimodal learning, and they fail to fully exploit the synergistic effect of multiscale semantics in answer reasoning. In this article, we propose a novel hierarchical synergy-enhanced multimodal relational network (HMRNet) to address these issues. Specifically, we devise (i) a compact and unified relation-oriented interaction module that explores the relation between intra- and inter-modal interactions to enable effective multimodal learning; and (ii) a hierarchical synergistic memory unit that leverages a memory-based interaction scheme to complement and fuse multimodal semantics at multiple scales to achieve synergistic enhancement of answer reasoning. With careful design of each component, our HMRNet has fewer parameters and is computationally efficient. Extensive experiments and qualitative analyses demonstrate that the HMRNet is superior to previous state-of-the-art methods on eight benchmark datasets. We also demonstrate the effectiveness of the different components of our method. [ABSTRACT FROM AUTHOR]

Regression-based face alignment involves learning a series of mapping functions to predict the true landmarks from an initial estimation of the alignment. Most existing approaches focus on learning efficacious mapping functions from some feature representations to improve performance. The issues related to the initial alignment estimation and the final learning objective, however, receive less attention. This work proposes a deep regression architecture with progressive reinitialization and a new error-driven learning loss function to explicitly address the above two issues. Given an image with a rough face detection result, the full face region is first mapped by a supervised spatial transformer network to a normalized form and trained to regress coarse positions of landmarks. Then, different face parts are further respectively reinitialized to their own normalized states, followed by another regression sub-network to refine the landmark positions. To deal with the inconsistent annotations in existing training datasets, we further propose an adaptive landmark-weighted loss function. It dynamically adjusts the importance of different landmarks according to their learning errors during training without depending on any hyper-parameters manually set by trial and error. A high level of robustness to annotation inconsistencies is thus achieved. The whole deep architecture permits training from end to end, and extensive experimental analyses and comparisons demonstrate its effectiveness and efficiency. The source code, trained models, and experimental results are made available at https://github.com/shaoxiaohu/Face_Alignment_DPR.git. [ABSTRACT FROM AUTHOR]

Abstract Background Previous studies have revealed that the lysin motif (LysM) domains of bacterial cell wall-degrading enzymes are able to bind to peptidoglycan moieties of the cell wall. This suggests an approach for a cell surface display system in Gram-positive bacteria using a LysM-containing protein as the anchoring motif. In this study, we developed a new surface display system in B. thuringiensis using a LysM-containing peptidoglycan hydrolase, endo-β-N-acetylglucosaminidase (Mbg), as the anchor protein. Results Homology searching in the B. thuringiensis YBT-1520 genome revealed a putative peptidoglycan hydrolase gene. The encoded protein, Mbg, exhibited substantial cell-wall binding capacity. The deduced amino acid sequence of Mbg was structurally distinguished as an N-terminal domain with two tandemly aligned LysMs and a C-terminal catalytic domain. A GFP-fusion protein was expressed and used to verify the surface localization by Western blot, flow cytometry, protease accessibility, SDS sensitivity, immunofluorescence, and electron microscopy assays. Low-level constitutive expression of Mbg was elevated by introducing a sporulation-independent promoter of cry3Aa. Truncated Mbg domains with separate N-terminus (Mbgn), C-terminus (Mbgc), LysM1, or LysM2 were further compared for their cell-wall displaying efficiencies. The Mbgn moiety contributed to cell-wall anchoring, while LysM1 was the active domain. Two tandemly repeated Mbgns exhibited the highest display activity, while the activity of three repeated Mbgns was decreased. A heterologous bacterial multicopper oxidase (WlacD) was successfully displayed onto the surface of B. thuringiensis target cells using the optimum (Mbgn)2 anchor, without radically altering its catalytic activity. Conclusion Mbg can be a functional anchor protein to target different heterologous proteins onto the surface of B. thuringiensis cells. Since the LysM domain appears to be universal in Gram-positive bacteria, the strategy presented here could be applicable in other bacteria for developing this type of system.

Bacillus thuringiensis has been widely used as a biopesticide for a long time. Here we report the finished and annotated genome sequence of B. thuringiensis mutant strain BMB171, an acrystalliferous mutant strain with a high transformation frequency obtained and stocked in our laboratory. doi: 10.1128/JB.00562-10

In this report, we utilized N-Acetylmuraminidase (AcmA) to develop a whole-cell catalyst of endo-beta-1, 3-1, 4-glucanase in Lactococcus lactis. The PCR-amplified full-length acmA gene from L. lactis MB191 was fused with the green fluorescent gene (gfp), followed by ligating the chimeric acmA-gfp into the Escherichia coli-L. lactis shuttle expression vector pMG36k, yielding the recombinant plasmid pMB137. SDS-PAGE analysis showed that the constitutive expression of AcmA-GFP fusion protein in the L. lactis AS1.2829 construct harboring pMB137 (named MB137), with the predicted Mr of 74 kD. Western blotting, GFP specific fluorescence intensity assays and flow cytometry analysis confirmed that AcmA-GFP was immobilized on the outer membrane, which constituted approx. 35% of the total intracellular fusion protein. Furthermore, acmA was fused with a PCR-amplified encoding fragment of the endo-beta-1, 3-1, 4-glucanase gene (gls) from Bacillus sublitis BF7658, resulting in the recombinant plasmid pMB138. By transferring pMB138 into L. lactis AS1.2829, the derived L. lactis MB138 expressing the AcmA-GLS fusion enzyme exhibited a distinct whole-cell glucanase activity (by 12 U/mL) compared to the control strain, indicating AcmA had served as a functional anchoring motif to immobilize the heterologous enzyme on the cell surface of L. lactis.