Learning effective fusion of multi-modality features is at the heart of visual question answering. We propose a novel method of dynamically fusing multi-modal features with intra-and inter-modality information flow, which alternatively pass dynamic information between and across the visual and language modalities. It can robustly capture the high-level interactions between language and vision domains, thus significantly improves the performance of visual question answering. We also show that the proposed dynamic intra-modality attention flow conditioned on the other modality can dynamically modulate the intramodality attention of the target modality, which is vital for multimodality feature fusion. Experimental evaluations on the VQA 2.0 dataset show that the proposed method achieves state-of-the-art VQA performance. Extensive ablation studies are carried out for the comprehensive analysis of the proposed method.
This report presents our method which wins the nuScenes 3D Detection Challenge [17] held in Workshop on Autonomous Driving(WAD, CVPR 2019). Generally, we utilize sparse 3D convolution to extract rich semantic features, which are then fed into a class-balanced multi-head network to perform 3D object detection. To handle the severe class imbalance problem inherent in the autonomous driving scenarios, we design a class-balanced sampling and augmentation strategy to generate a more balanced data distribution. Furthermore, we propose a balanced grouping head to boost the performance for the categories with similar shapes. Based on the Challenge results, our method outperforms the PointPillars [14] baseline by a large margin across all metrics, achieving state-of-the-art (SOTA) detection performance on the nuScenes dataset. Code will be released at CBGS.
Mesoporous TiO 2 nanometer thin films were prepared on fused quartz by the dip-coating sol-gel method from a system containing a triblock copolymer as a template (or pore-forming agent), and then calcined at different temperatures. These films were characterized by X-ray diffraction, atomic force microscopy, X-ray photoelectron spectroscopy, BET surface area and UV-visible spectrophotometry. The photocatalytic activity and photo-induced super-hydrophilicity of the films were evaluated by the photocatalytic degradation of acetone and water contact angle measurement in air, respectively. It was found that the thin films calcined at 700 C not only show the highest photocatalytic activity, but also possess the greatest light-induced hydrophilicity and the slowest conversion rate from the hydrophilic to a hydrophobic state. The former is attributed to the fact that the films calcined at 700 C are composed of anatase and rutile, whichis beneficial in enhancing the transfer of photo-generated electrons from the anatase to the rutile phase, reducing the electronhole combination rate in anatase and enhancing its activity. The high light-induced hydrophilicity and slow hydrophilic to hydrophobic conversion rate are due to the synergetic effect of good photocatalytic activity, sufficient surface hydroxyl content and a degree of surface roughness. Because of their high specific surface areas and mesoporous structures, the photocatalytic activity of mesoporous TiO 2 thin films is higher than that of conventional TiO 2 thin films.Porous materials have attracted more and more attention in the fields of separation and catalysis owing to their high specific areas and pore volumes, as well as their narrow pore size distributions. Mesoporous materials were first synthesized by Mobil researchers in 1992, 1,2 and have played an important role in both basic and applied research areas in recent years. Many metal oxides, especially porous ones, have special catalytic functions. 3-8 Titania has been widely used as a catalyst in organic reactions such as selective oxidation and photocatalytic reactions such as alcohol dehydration, photo-Kolbe oxidations of organic acids, 9 oxidation of aromatic compounds, 10 degradation of paint pigments 11 and nitrogen oxide reduction. 12 Mesoporous titania should be an even more effective catalyst because of its large surface area and porosity. Mesoporous titania (Ti-TMS1) was first synthesized through modified sol-gel routes in the presence of alklyphosphate surfactant templates by Antonelli and Ying in 1995. 5 Since then, numerous research papers on synthetic methods and catalytic properties have appeared. [6][7][8][13][14][15][16][17][18][19] Recent investigations have confirmed that block polymers could also be used as templates to direct the formation of mesoporous titania and other oxides. 6-8 Very recently, we have also reported methods for the preparation of mesoporous TiO 2 thin films and powders by reverse micelle synthesis and ultrasonic illumination. [20][21][22][23] In the present work, we ...
High performance liquid chromatography coupled with tandem mass spectrometry with electrospray ionization and negative ion detection (HPLC-ESI-MS/MS) was used to identify the polyphenols from the leaves of a traditional vegetable and spice, Zanthoxylum bungeanum Maxim., in China by comparison of the retention times and mass spectral fragment with standard substances or related literature. A phenolic acid glucoside, two acids, and eight flavonoids were determined including vanillic acid-4-glucoside (22.75 μg/g, dry weight of leaves), quinic acid (58.58 μg/g), chlorogenic acid (2515.96 μg/g), epicatechin (77.80 μg/g), 5-feruloyquinic acid (16.63 μg/g), syringetin-3-glucoside (103.23 μg/g), rutin (89.41 μg/g), hyperoside (886.36 μg/g), quercetin-3-arabinoside (118.75 μg/g), quercitrin (645.82 μg/g), and isorhamnetin-3-glucoside (104.27 μg/g). The obtained polyphenols show good radical scavenging activities in DPPH and superoxide anion. Through the intracellular antioxidation assay, the result shows that the polyphenols could reduce the ROS in HT-29 cells without cell toxicity.
The essential oil of the seeds of Cuminum cyminum L. from China was isolated by hydrodistillation in a yield of 3.8%. The chemical composition of the essential oil was examined by GC and GC-MS; 37 components, representing 97.97% of the oil, were identified. Cuminal (36.31%), cuminic alcohol (16.92%), γ γ γ γ γ -terpinene (11.14%), safranal (10.87%), p-cymene (9.85%) and β β β β β-pinene (7.75%) were the major components.
Porous titania microspheres with a very narrow particle size distribution (PSD) were synthesized by polymerization-induced colloid aggregation (PICA). After being sintered, the titania microspheres that are obtained have an average diameter of 3.5 microm, a surface area of 36.7 m2/g, an average pore volume of 0.30 mL/g, and an average pore diameter of 32.2 nm. The microspheres possess enough rigidity to withstand high packing pressure and are very useful as a new kind of chromatographic packing material for high performance liquid chromatography (HPLC).
Flavonoids are rich in Juglans regia L. leaves. They have potent antioxidant properties, which have been related to regulating immune function and enhancing anticancer activity. Herein, qualitative and quantitative determination of flavonoids from J. regia leaves was carried out using high performance liquid chromatography coupled with tandem mass spectrometry with electrospray ionization and negative ion detection (HPLC-ESI-MS/MS) by comparison of the retention times and mass spectral fragments with standard substances or related literatures. Seventeen compounds were identified and major components are quercetin-3-O-rhamnoside (453.11 μg/g, dry weight), quercetin-3-O-arabinoside (73.91 μg/g), quercetin-3-O-xyloside (70.04 μg/g), kaempferol-O-pentoside derivative (49.04 μg/g), quercetin-3-O-galactoside (48.61 μg/g), and kaempferol-O-pentoside (48.46 μg/g). The in vitro intracellular antioxidation indicated that flavonoids from J. regia leaves could reduce the reactive oxygen species (ROS) level in RAW264.7 cells and showed good radical scavenging activities. These results proved to be more related to the flavonoids that could be considered in the design of new formulations of dietary supplements or functional foods.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.
hi@scite.ai
10624 S. Eastern Ave., Ste. A-614
Henderson, NV 89052, USA
Copyright © 2024 scite LLC. All rights reserved.
Made with 💙 for researchers
Part of the Research Solutions Family.