Dynamic hand gesture recognition has attracted increasing attention because of its importance for human⁻computer interaction. In this paper, we propose a novel motion feature augmented network (MFA-Net) dynamic from skeletal data. MFA-Net exploits features finger and global movements to augment deep recognition. To describe articulated movements, are extracted the skeleton sequence via variational autoencoder. Global utilized represent skeleton. These along with then fed into three branches...

Abstract Mass spectrometry (MS) promises small‐metabolite profiling as a tool of the future and calls for comprehensive understanding key procedures to enhance its capability. Herein, we studied cation adduction fragmentation small metabolites by combination theoretical experimental approaches, via nanoparticle‐assisted laser desorption/ionization (LDI)‐MS MS/MS. We calculated energies conformers atomic bond orders establish rules cation–metabolite affinity multiple adductions in charge...

This paper studies unsupervised/self-supervised whole-graph representation learning, which is critical in many tasks such as molecule properties prediction drug and material discovery. Existing methods mainly focus on preserving the local similarity structure between different graph instances but fail to discover global semantic of entire data set. In this paper, we propose a unified framework called Local-instance Global-semantic Learning (GraphLoG) for self-supervised learning....

We develop a novel self-supervised learning method named Shape Self-Correction for point cloud analysis. Our is motivated by the principle that good shape representation should be able to find distorted parts of and correct them. To learn strong representations in an unsupervised manner, we first design shape-disorganizing module destroy certain local object. Then destroyed normal are sent into network get representations, which employed segment points belong further reconstruct them restore...

Autonomous highlight detection is crucial for enhancing the efficiency of video browsing on social media platforms. To attain this goal in a data-driven way, one may often face situation where annotations are not available target category used practice, while supervision another (named as source category) achievable. In such situation, can derive an effective detector by transferring knowledge acquired from to one. We call problem cross-category detection, which has been rarely studied...

The WOX (WUSCHEL-related homeobox) gene family plays pivotal roles in plant growth, development, and responses to biotic/abiotic stresses. Flax (Linum usitatissimum L.), a globally important oilseed fiber crop, lacks comprehensive characterization of its family. Here, 18 LuWOX genes were systematically identified the flax genome through bioinformatics analyses. Phylogenetic classification grouped these into three clades: Ancient, Intermediate, WUS Clades, with members within same clade...

Multi-Source Domain Adaptation (MSDA) focuses on transferring the knowledge from multiple source domains to target domain, which is a more practical and challenging problem compared conventional single-source domain adaptation. In this problem, it essential model jointly, an effective combination scheme also highly required. The graphical structure among different useful tackle these challenges, in interdependency various instances/categories can be effectively modeled. work, we propose two...

Face identity editing (FIE) shows great value in AI content creation. Low-resolution FIE approaches have achieved tremendous progress, but high-quality struggles. Two major challenges hinder higher-resolution and higher-performance development of FIE: lack high-resolution dataset unacceptable complexity forbidding for mobile platforms. To address both issues, we establish a novel large-scale, tailored FIE. Based on our SimSwap (Chen et al. 2020), propose an upgraded version named SimSwap++...

Mass spectrometry has been applied to the targeted analysis of commonly used additives (such as Irganox 1010, 1076, Irgafos 168 etc.) in plastic materials, but a fast and straightforward method for non-targeted identification quantification unusual or potentially new antioxidant is still unavailable. In this study, novel simple unknown food packaging using ultrasonic extraction ultra-performance liquid chromatography coupled with quadrupole time-of-flight mass developed. A series analyzed...

Recently, a backdoor data poisoning attack was proposed, which adds mislabeled examples to the training set, with an embedded pattern, aiming have classifier learn classify target class whenever pattern is present in test sample. We address post-training detection of innocuous perceptible backdoors DNN image classifiers, wherein defender does not access poisoned set. This problem challenging because without we no hint about actual used during training. identify two properties patterns -...

Abstract In the issue of few-shot image classification, due to lack sufficient data, directly training model will lead overfitting. order alleviate this problem, more and methods focus on non-parametric data augmentation, which uses information known construct normal distribution expand samples in support set. However, there are some differences between base class new ones, different belonging same is also different. The sample features generated by current may have deviations. A...

It is difficult to identify unknown impurities in nucleotide analogues by mass spectrometry because mass‐spectrometry‐incompatible mobile phases need be used separate the major ingredient from impurities. In this study, vidarabine monophosphate was selected, and were identified online heart‐cutting two‐dimensional high‐performance liquid chromatography linear ion trap spectrometry. The one‐dimensional reversed‐phase column filled with a phase containing nonvolatile salt. chromatography, we...

The contrastive vision-language pre-training, known as CLIP, demonstrates remarkable potential in perceiving open-world visual concepts, enabling effective zero-shot image recognition. Nevertheless, few-shot learning methods based on CLIP typically require offline fine-tuning of the parameters samples, resulting longer inference time and risk over-fitting certain domains. To tackle these challenges, we propose Meta-Adapter, a lightweight residual-style adapter, to refine features guided by...

We address cross-species 3D face morphing (i.e., from human to animal), a novel problem with promising applications in social media and movie industry. It remains challenging how preserve target structural information source fine-grained facial details simultaneously. To this end, we propose an Alignment-aware Face Morphing (AFM) framework, which builds semantic-adaptive correspondence between faces across species, via alignment-aware controller mesh (Explicit Controller, EC) explicit...

Abstract Mass spectrometry (MS) promises small‐metabolite profiling as a tool of the future and calls for comprehensive understanding key procedures to enhance its capability. Herein, we studied cation adduction fragmentation small metabolites by combination theoretical experimental approaches, via nanoparticle‐assisted laser desorption/ionization (LDI)‐MS MS/MS. We calculated energies conformers atomic bond orders establish rules cation–metabolite affinity multiple adductions in charge...