Search

We introduce MQ-Det, an efficient architecture and pre-training strategy design to utilize both textual description with open-set generalization and visual exemplars with rich description granularity as category queries, namely, Multi-modal Queried object Detection, for real-world detection with both open-vocabulary categories and various granularity. MQ-Det incorporates vision queries into existing well-established language-queried-only detectors. A plug-and-play gated class-scalable perceiver module upon the frozen detector is proposed to augment category text with class-wise visual information. To address the learning inertia problem brought by the frozen detector, a vision conditioned masked language prediction strategy is proposed. MQ-Det's simple yet effective architecture and training strategy design is compatible with most language-queried object detectors, thus yielding versatile applications. Experimental results demonstrate that multi-modal queries largely boost open-world detection. For instance, MQ-Det significantly improves the state-of-the-art open-set detector GLIP by +7.8% zero-shot AP on the LVIS benchmark and averagely +6.3% AP on 13 few-shot downstream tasks, with merely 3% pre-training time required by GLIP. Code is available at https://github.com/YifanXu74/MQ-Det., Comment: Under review

Although significant progress has been made in few-shot learning, most of existing few-shot image classification methods require supervised pre-training on a large amount of samples of base classes, which limits their generalization ability in real world application. Recently, large-scale Vision-Language Pre-trained models (VLPs) have been gaining increasing attention in few-shot learning because they can provide a new paradigm for transferable visual representation learning with easily available text on the Web. However, the VLPs may neglect detailed visual information that is difficult to describe by language sentences, but important for learning an effective classifier to distinguish different images. To address the above problem, we propose a new framework, named Semantic-guided Visual Adapting (SgVA), which can effectively extend vision-language pre-trained models to produce discriminative adapted visual features by comprehensively using an implicit knowledge distillation, a vision-specific contrastive loss, and a cross-modal contrastive loss. The implicit knowledge distillation is designed to transfer the fine-grained cross-modal knowledge to guide the updating of the vision adapter. State-of-the-art results on 13 datasets demonstrate that the adapted visual features can well complement the cross-modal features to improve few-shot image classification.

Microstructure and mechanical properties of Bypass-Current Metal Inert Gas (BC-MIG) welded AA6061/Ti6Al4V lap joint were studied. Due to the influence of bypass current, the heat input acting on base metal was reduced, the melting efficiency of filler wire was increased, and the metal transfer mode was changed from short circuiting transfer to projected transfer, which creates an improvement in wettability and spreading behavior of weld. Reliable Al/Ti joints were obtained by BC-MIG welding-brazing, and the tensile-shear strength of as-received joint could reach a maximum of 190 M Pa, which was about 96% of AA6061 base metal. The metallurgical bonding between Al alloy and Ti alloy was a continuous reaction layer, and the intermetallic compounds (IMCs) consisted of TiAl3 and TiAl. The formation and growth of interface were discussed by melt flow.

Graph convolutional networks (GCNs) based methods have achieved advanced performance on skeleton-based action recognition task. However, the skeleton graph cannot fully represent the motion information contained in skeleton data. In addition, the topology of the skeleton graph in the GCN-based methods is manually set according to natural connections, and it is fixed for all samples, which cannot well adapt to different situations. In this work, we propose a novel dynamic hypergraph convolutional networks (DHGCN) for skeleton-based action recognition. DHGCN uses hypergraph to represent the skeleton structure to effectively exploit the motion information contained in human joints. Each joint in the skeleton hypergraph is dynamically assigned the corresponding weight according to its moving, and the hypergraph topology in our model can be dynamically adjusted to different samples according to the relationship between the joints. Experimental results demonstrate that the performance of our model achieves competitive performance on three datasets: Kinetics-Skeleton 400, NTU RGB+D 60, and NTU RGB+D 120., Comment: 12 pages, 6 figures

Additional file 1: Table S1. Gene-specific primers for the amplification of the CHSA cDNA from Culex pipiens pallens. Table S2. Gene-specific primers for qPCR. qPCR, quantitative real-time PCR. Table S3. Gene-specific primers for RNAi. RNAi, RNA interference; siRNA, small interfering RNA. Table S4. CpCHSA sequences used for polyclonal antibody preparation.

The solid state drive based on NAND flash memory has been widely used, but the limited number of Program/Erase operation cycles has lead to its limited lifetime and reliability, reduced the endurance of solid state drive. The lifetime of the NAND flash-based solid state drive depends on its endurance, and the write amplification is an important factor that affects the endurance. It is of great value to test and evaluate the write amplification in real and accurate. The accuracy and credibility of the evaluation model are seldom mentioned in the existing research, and some characteristics of the solid state drive are neglected, therefore, the simulation model based on solid state drive has a credibility problem. In this paper, the actual volume data of programming operation in NAND flash memory chip is obtained by detecting the change signal of the pin on the flash chip, flash memory chip write amplification is tested and quantified, solid state drive Program/Erase wastage evaluation model based on NAND flash memory is studied and established, namely the endurance evaluation model (EEM). The experimental results show that the EEM is feasible and can be used to evaluate the endurance of solid state drive.

In the era of big data, with the rapid expansion of information, large-scale storage system has been widely used. However, the problem of construction cost and energy waste due to low storage utilization has been more obvious in large-scale storage systems, also increased the difficulty of information management and maintenance investment, therefore, big data storage technologies such as storage virtualization, data deduplication and others appear to improve storage utilization. At present, these techniques have been widely used, but so far there is no effective indicators, methods and tools to evaluate its effectiveness. In this paper, storage virtualization is analyzed, and the evaluation indicators and evaluation method, and the design and implementation of the evaluation tool are described. The experimental results show that the evaluation indicators, method and tool are practical, it further provides a new method for quantitative evaluation of the storage utilization of large-scale storage system, helps enterprises better to plan storage system, and reduce the waste of resources and improve the rate of return on investment.

According to the actual situation of sweet potato industry development in Ziyun County, we use SWOT analysis method to conduct strategic analysis on strengths and weaknesses of the internal environment and the opportunities and challenges of the external environment of the sweet potato industry, to explore correct strategic countermeasures suitable for the future development of sweet potato industry in Ziyun County and provide reference for the sweet potato industry to maintain sustainable competitiveness.

建立了分散固相萃取-同位素稀释-气相色谱-串联质谱同时测定动物源性食品中9种 N -亚硝胺的方法。样品用乙腈提取，上清液经分散固相萃取（dSPE）净化后浓缩。选用DB-WAX极性毛细管色谱柱对待测物进行分离，经EI源电离后以多反应监测（MRM）模式采集数据并做定性筛查和定量分析。9种 N -亚硝胺在相应的浓度范围内线性关系良好，相关系数均大于0.99，对4类典型的动物源性食品进行3个不同浓度的加标试验，平均回收率为62.5%~118%，RSD为2.11%~25.6%，检出限（LOD， S/N =3）为0.02~0.31 μg/kg。该方法成本低廉、灵敏可靠，适用于同时对动物源性食品中9种 N -亚硝胺进行定性和定量测定。