一种基于Involution卷积的三维人体重建方法

针对从视频中恢复三维人体模型运动序列时,由于图像特征提取能力有限而导致三维人体模型运动序列重建效果不佳的问题,提出了一种基于Involution卷积的三维人体重建方法。首先为了引入自注意力机制,在ResNet50网络结构中加入Involution算子,获取视频图像帧的特征向量,然后使用姿态估计网络和形状估计网络获取人体姿势以及形状参数,最后使用蒙皮多人线性模型(skinned multi-person linear model, SMPL)生成三维人体模型的运动序列。在三维姿态户外数据集(3D pose in the wild, 3DPW)上与视频人体姿态形状估计推理(video inference for body pose and shape estimation, VIBE)方法以及时间一致性网格恢复(temporally consistent mesh recovery, TCMR)方法进行对比实验,平均精度相比于VIBE、TCMR分别提升了3.1%、0.7%,能够为运动捕捉、三维人体动画制作等工作提供更为准确的三维人体模型。

GRU和Involution改进的深度伪造视频检测方法

近年来深度伪造视频在网络上广泛传播,造成了负面影响。针对现有检测模型准确率低和信息提取不够充分和全面的问题,提出了一种GRU(gated recurrent unit)和Involution改进的深度伪造视频检测方法。首先使用VGG19作为主干网络提取空间特征,并将Involution算子嵌入主干网络,从空间和通道信息两方面加强了人脸图像的空间建模能力。然后通过主胶囊层关注特征的位置信息和使用GRU提取帧间的时序特征。最后在训练模型阶段使用focalloss作为损失函数来平衡样本。在Deepfakes、FaceSwap和Celeb-DF数据集中进行测试,实验结果表明该方法优于主流检测方法,改进对比实验进一步证明了检测方法的有效性。

注意力机制与Involution算子改进的人脸表情识别

针对复杂人脸表情识别面临背景干扰、空间信息分布不均匀等问题,提出一种注意力机制和Involution算子改进的人脸表情识别方法,该方法以VGG19为基线网络,前端引入注意力机制提取表情强相关特征,抑制背景干扰,并利用联合正则化策略平衡和改善特征数据分布,提高模型训练质量;后端采用密集连接加强有效特征复用,提取高层语义信息。所提方法在CK+、FER2013、RAF-DB等3个公开数据集上进行了验证,准确率均取得显著提高,且优于当前诸多先进方法。此外,为提高网络处理复杂条件下的数据集,在其后端引入Involution算子替代部分卷积层,提高了空间多样性信息学习能力。实验结果表明,所提模型可有效提高RAF-DB等复杂数据集的人脸表情识别准确率。

基于Involution算子和交叉注意力机制的合成语音检测方法

随着科学技术的迅速发展,基于深度学习生成的合成语音给语音认证系统和网络空间安全带来了新的挑战。针对现有检测模型准确率较低和语音特征挖掘不够充分的问题,提出了一种基于Involution算子和交叉注意力机制改进的合成语音检测方法。前端将语音数据提取线性频率倒谱系数(LFCC)特征和恒定Q变换(CQT)谱图特征,两个特征分别输入到后端的双分支网络中。后端网络使用ResNet18作为主干网络先进行浅层的特征学习,并将Involution算子嵌入主干网络,扩大特征图像学习区域,增强在空间范围内学习到的频谱图像特征信息。同时在训练分支之后引入cross-attention交叉注意力机制,使LFCC特征和CQT谱图特征构建交互的全局信息,强化模型对特征的深层挖掘。所提模型在ASVspoof 2019 LA测试集上取得了0.84%的等错误率和0.026的最小归一化串联检测代价函数的实验结果,展现了优于主流的检测模型。结果表明,改进的模型能够有效融合不同的频谱特征,提高模型的特征学习能力,从而强化模型的检测能力。

一种基于Involution和BERT词向量的情感分析算法

文本情感分析是当前数据分析和数据挖掘的一个重要的研究方向,提出了一种融合Involution和BERT词向量的情感分析算法,首先,通过对预处理之后的数据传入BERT模块,用学习质量更高的文本数据词向量表示;其次,利用Involution核函数方法进行更好的特征分析和提取;最后,在三种标准数据集上进行了广泛的实验,研究结果表明,基于Involution和BERT的情感分析模型能够有效地提取文本数据中的有用信息,并准确地进行情感分析,从而更好地了解用户的情感偏好.

Renovation and Involution — Paradox of Chinese Literature in Modernity

With a special matter, Chinese literature has differences with traditional theories from western countries in essence, so that Chinese literature demonstrates its renovation and involution. The main reason is its trends of modernity. The transformation from oversea exchange is very important. This paper mainly explains the specific content of paradox of Chinese literature in mo dernity, and explores the reasons for differences between China and western countries.

基于小波变换和Involution卷积神经网络的滚动轴承故障诊断方法

传统卷积神经网络在滚动轴承故障诊断中存在梯度扩散、参数爆炸和训练时间长等缺点,为此提出一种基于Involution卷积神经网络的滚动轴承故障诊断方法。首先,通过小波变换将原始数据转换为时频图,经过数据预处理后送入基于Involution的神经网络模型;然后,经过不同Convolution卷积核对原始图像的局部区域特征进行不同方式的提取,扩展输入图像通道数;最后,通过Involution对特征图的每个像素点进行进一步特征提取,经过softmax层输出分类结果。CWRU轴承数据集和试验轴承数据集的分析结果表明,基于Involution卷积的神经网络模型所包含的参数较少,训练时间短,故障分类准确率可达99.75%,优于传统的CNN,DBN,自编码器等模型。

INVOLUTIONS WITH FIXED POINT SET RP(2m)∪P(2m, 2n + 1)

Let （M, T） be a smooth closed manifold with a smooth involution T whose fixed point set is a disjoint union of an even-dimensional real projective space and a Dold manifold. In some cases, the equivariant bordism classes of （M, T） are determined.

INVOLUTIONS FIXING ∪（from i=1 to m） CP_i(1) × HP_i(n)

Let （M, T） be a closed manifold with an involution T. The fixed point set of T is F. In this article, bordism classes of the involutions with fixed point set F = ^mUi=1 CPi（1）×HPi（n） are determined, where CP（1） and HP（n） denote the 1-dimensional i=1 complex projective space and n-dimensional quaternionic projective space respectively, and n=2^p-2or n=2^p-1（p〉 1）.

“Involution”与“内卷”概念的翻译对等度考察

“内卷”作为一个从英语世界译入的外来词,1986年进入国内学术界,而在2020~2021年经过中国网络社区发酵后,中国官方在英语世界解释的“内卷”与1986的译入的“内卷”已意义悬殊。“involution”在中译后获得了“内卷”这一能指符号,该能指符号虽用于经济学语境,但通过中文语料库的研究显示,在大众语境下主要还是反映植物学界的世界观和理论概念。而在2020年“内卷”这一能指挣脱原来的所指范畴,产生了“内耗式内部竞争”这一新的社会心理学所指意义。这一跨文化旅行所带来的符号意义的增加,是理论旅行中文化语境参数的变化带来的概念变异,也是模因传播模仿和复制时产生的语义迁徙。出于经济原则,“involution”与“内卷”直接构成了主流媒体上的对等词,但其对等关系需要进行二度解释才能被西方世界所理解和接受。

SOME PROPERTIES OF COMMUTING AND ANTI-COMMUTING m-INVOLUTIONS

We define an m-involution to be a matrix K ∈ Cn×n for which Km -= I. In this article, we investigate the class Sm （A） of m-involutions that commute with a diagonalizable matrix A E Cn×n. A number of basic properties of Sm （A） and its related subclass Sm （A, X） are given, where X is an eigenvector matrix of A. Among them, Sm （A） is shown to have a torsion group structure under matrix multiplication if A has distinct eigenvalues and has non-denumerable cardinality otherwise. The constructive definition of Sm （A, X） allows one to generate all m-involutions commuting with a matrix with distinct eigenvalues. Some related results are also given for the class S,, （A） of m-involutions that anti-commute with a matrix A ∈ Cnn×n.

Regulation of leptin in involution of mammary gland

Leptin, a protein hormone produced and secreted predominantly by white adipose tissue, has a critical role in the regulation and coordination of energy metabolism. Leptin is produced in the mammary gland by the fat tissue or by the mammary epithelium. In vitro study has shown that leptin triggers apoptosis in mammary epithelial cells. Mammary gland involution is characterized by extensive apoptosis of the epithelial cells. At the onset of involution, STAT3 is specifically activated.Various studies show that leptin act as a paracrine and autocrin factor to influence mammary epithelial cell proliferation and differentiation. This paper reviewed the function of leptin to the involution of mammary gland.

Relationship among uterine involution, ovarian activity, blood metabolites and subsequent reproductive performance in Egyptian buffaloes

The aim of the study to monitor post partal uterine involution, ovarian activity and biochemical parameters and it is relation to the subsequent fertility. A total sixty normal calving pluriparous buffaloes were examined between 14th and 75th day post partum (p.p.) rectaly, ultrasonically and blood sampling were collected on weekly sessions. There were differences (P < 0.01) between pregnant (PREG) and non-pregnant (NPREG) groups in Body condition score (BCS) and body weights. There was a difference between previous gravid uterine horn (PGUH) and non-gravid uterine horn (NPGUH) diameter in PREG and NPREG groups at 28th day p.p. The calving to first service interval in the PREG group was shorter (P = 0.03) than that of NPREG one. The number of buffaloes with dominant follicles (DF 3 8 mmdiameter) in ipsilateral and contralateral ovary to the PGUH in PREG group was higher (P < 0.01) than in NPREG. The calving to first service interval in the PREG group having DF in the ovary ipsilateral to the PGUH (n = 16) was shorter (P < 0.01) than those buffaloes having no DF (n = 18). The number of service per conception and days open in the PREG buffaloes which had no DF in the ovary ipsilateral to the PGUH were higher and longer (P < 0.01) than that which had DF group. The values of glucose and triglyceride were higher (P = 0.057) in PREG than NPREG group. In conclusion, postpartum ovarian activity has positive effect on the uterine involution and postpartum profile of some metabolities may be a good predictor of fertility status of buffaloes.

The Effectiveness of Postpartum Exercise and Oxytocin Massage on Uterus Involution

The process of involution will work well if uterus contraction is strong, so a treatment can be done to improve uterus contraction. Postpartum exercise can make uterus muscles undergo contractions and retractions. In addition to pospartum exercise, being control of bleeding from the placenta attached is maintaining contraction and retraction fiber of strong myometrium through oxytocin massage. Purpose: This study aimed to ascertain influences of postpartum exercise and oxytocin massage on uterus involution. Method: The design used is quasi experimental research using post test only designing two comparison treatments. This research was conducted at PONED (basic emergency obstetric and neonatal care) in Bandung. The sample size was 50 respondents (25 respondents were given treatment for postpartum exercise and 25 respondents were given the oxytocin massage treatment). The sampling technique with non probability sampling techniques was by means of quota sampling. The data use primary data through the process of interview and observation. Deep data collection tool uses observation sheets and questionnaires. Result: The average uterus involution on the postpartum exercise group was 8.68 days, while the oxytocin massage group was 6.72 days. The results of the Mann Whitney Test statistic were at 95% confidence level;there was a significant difference in uterus involution between the postpartum exercise group and the oxytocin massage group with a value of p Conclusion: Oxytocin massage is more influential than pospartum exercise to uterus involution.

Dual Channel with Involution for Long-Tailed Visual Recognition

With the rapid increase of large-scale problems, the distribution of real-world datasets tends to be long-tailed. Existing solutions typically involve re-balancing strategies (i.e., re-sampling and re-weighting). Although they can significantly promote the classifier learning of deep networks, they will unexpectedly impair the representative ability of the learned deep features to a certain extent. Therefore, this paper proposes a dual-channel learning algorithm with involution neural networks (DC-Invo) to take care of representation learning and classifier learning concurrently. In this work, the most important thing is to combine ResNet and involution to obtain higher classification accuracy because of involution’s wider coverage in the spatial dimension. The paper conducted extensive experiments on several benchmark vision tasks including Cifar-LT, Imagenet-LT, and Places-LT, showing that DC-Invo is able to achieve significant performance gained on long-tailed datasets.

Jordan Γ*-Derivation on Semiprime Γ-Ring M with Involution

Let M be a 2-torsion free semiprime G-ring with involution satisfying the condition that ( and ). In this paper, we will prove that if a non-zero Jordan G*-derivation d on M satisfies for all and , then .

内卷还是合转——诠释学视域下的Involution理解框架论辩及转换

作为involution中译词的“内卷”,在当下国内网络流行话语中是一个充斥着负面情绪的热词。囿于分析模型的合法性危机及语用学困境,“内卷”被用于联系involution与中国社会发展的中介词作用即将消解。在社会—历史生产翻译(诠释)观视域中,通过一种形式批判和社会—历史诠释可以发现,“合转”有可能替代“内卷”成为联系involution与中国社会发展的新型中介词。在澄清其形式与社会—历史逻辑结构基础上,“合转”作为involution的新译词,可被用于理解当下中国社会发展的知识合法性,通过反观、反思和追问来寻找符合本土实际的真正创新之路,以推进中国式现代化的发展和完善。

基于块注意力机制和Involution的文本情感分析模型

【目的】解决卷积核宽度与词向量维度相同使得卷积层参数过多的问题,解决卷积操作的稀疏连接以及卷积的空间不变性和通道特异性不适用于文本任务的问题。【方法】提出一种基于块注意力机制和Involution的文本情感分析模型。模型先对分词后的单个词向量进行变形,将一维词向量变形为n×n词矩阵块,然后将句子中多个词的词矩阵块拼接成句子矩阵。句子矩阵经过块注意力机制层,增强了文本特征的上下文相关性及位序信息,再通过采用具有空间特异性和通道不变性的Involution对句子矩阵进行特征提取,最后使用全连接层进行文本情感分类。【结果】在三个文本情感分析公开数据集waimai_10k、IMDB、Tweet上的实验表明,所提模型的分类准确率分别达到88.47%、86.22%、94.42%,与词向量卷积网络和循环神经网络中的Bi-LSTM模型相比准确率分别提高6.47、7.72、9.35个百分点和1.07、1.01、0.59个百分点。【局限】所提模型在大型数据集上的分类准确度低于中小型数据集。【结论】引入块注意力机制和Involution的文本情感分析模型解决了参数量过多、卷积操作的稀疏连接以及卷积的空间不变性和通道特异性的问题,在不同数据集上,与传统卷积模型比较,本文模型的准确率有所提升。

基于改进YOLOv7-tiny的高空作业人员安防装备检测算法

针对现有高空作业人员安防装备检测算法参数量较大且检测目标相对单一,难以适应复杂的高空作业场景等问题,提出一种基于改进YOLOv7-tiny的高空作业人员安防装备检测算法。首先,将主干网络重新设计为更轻量的YOLOv7-FasterNet,并调整空间金字塔池化结构,实现模型参数量的压缩;其次,在ELAN-L模块中扩展梯度传输路径的分支,解决了模型压缩造成的通道信息缺失问题,提升了特征信息的提取能力;最后,将网络中下采样部分替换为Involution模块,降低参数冗余,增强网络对全局的捕获能力。实验结果表明,改进的YOLOv7-tiny算法能够更好地适应复杂高空作业场景,在开源数据集上具备良好的性能。该算法的平均检测精度达到94.7%,较原模型提升1.5%,参数量较原模型下降11.6%,实验结果验证了算法改进措施的有效性。

Involution改进的卷积神经网络人群计数方法

针对现有人群计数方法大多采用卷积操作提取特征,空间多样性特征信息提取和传递能力不足的问题,提出一种Involution改进的单列深层人群计数网络。该网络以VGG-16为基本框架,引入Involution算子替代卷积操作,并辅以残差链接提高对空间特征信息的感知和传递能力;采用膨胀卷积保持分辨率的同时扩大感受野,丰富深度语义特征;利用联合损失函数监督网络训练,提高计数准确性和全局信息相关性。所提方法在公开数据集ShangHaiTech、UCF-QNRF、UCF_CC_50上的性能较基线模型提升显著,并超越了诸多当前的先进算法。结果表明:所提人群计数方法具有较高的准确性和更好的鲁棒性。