深度神经网络在静态图像领域已取得突破性进展,并逐步扩展到视频识别领域。人体动作识别是视频识别领域的研究热点和难点,因此,提出了一种基于双流快速区域卷积神经网络(Faster RCNN)改进的人体动作识别算法。首先,用RGB(Red,Green,Blue)图像和光流数据作为网络的输入,分别训练Faster RCNN;然后,将训练好后的网络模型进行融合,并引入改进的压缩和激励模块对特征通道进行处理,以突出重要特征;最后,用完全的交并比损失函数作为边框回归损失函数,以优化某些预测框与真实框不能相交等问题。实验结果表明,相比传统的Faster RCNN,本算法在动作识别数据集UCF101上的准确率得到了一定的提高。
2022-05-05 10:24:52
2.69MB
1
Redis流队列
Redis流v5 +上的消息队列。
redis v5 +流，ES6
__keyevent@5__:expired保证__keyevent@5__:expired消息（ __keyevent@5__:expired ）不丢失
采用ES6 +异步/等待+面向对象设计
达到易读懂，方便修改。
vscode上有专项预测开发的扩展插件：
安装
npm install redis-stream-queue
基本用法
const { RedisQueue } = require ( 'redis-stream-queue' )
const IORedis = require ( 'ioredis' )
const client = new IORedis ( opt )
const mq = RedisQueue . init ( { client } )
const sKey =
2022-04-21 19:01:23
19KB
1
当前卷积神经网络结构未能充分考虑RGB图像和深度图像的独立性和相关性, 针对其联合检测效率不高的问题, 提出了一种新的双流卷积网络。将RGB图像和深度图像分别输入到两个卷积网络中, 两个卷积网络结构相同且权值共享, 经过数次卷积提取各自独立的特征后, 在卷积层根据最优权值对两个卷积网络进行融合；继续使用卷积核提取融合后的特征, 最后通过全连接层得到输出。相比于以往卷积网络对RGB-D图像采用的早期融合和后期融合方法, 在检测时间相近的情况下, 双流卷积网络检测的准确率和成功率分别提高了4.1%和3.5%。
2022-03-10 01:04:52
9.69MB
1
时序行为检测是指在一段未分割的长视频中，检测出其中包含的若干行为片段的起止时间和类别.针对该项任务，提出基于双流卷积神经网络的行为检测模型.首先使用双流卷积神经网络提取视频的特征序列，然后使用TAG （Temporal Actionness Grouping）生成行为提议，为了构建高质量的行为提议，将行为提议送入边界回归网络中修正边界，使之更为贴近真实数据，再将行为提议扩展为含有上下文信息的三段式特征设计，最后使用多层感知机对行为进行识别.实验结果表明，本算法在THUMOS 2014数据集和ActivityNet v1.3数据集获得较好的识别率.
2021-12-26 15:25:55
1016KB
1
安装jnetpcap本地仓库
对于linux，sudo是先决条件
//linux :at the pathtoproject/jnetpcap/linux/jnetpcap-1.4.r1425
//windows: at the pathtoproject/jnetpcap/win/jnetpcap-1.4.r1425
mvn install:install-file -Dfile=jnetpcap.jar -DgroupId=org.jnetpcap -DartifactId=jnetpcap -Dversion=1.4.1 -Dpackaging=jar
跑步
IntelliJ IDEA
在IDE中打开终端
//linux:
$ sudo bash
$ ./gradlew execute
//windows:
$ gradlew execute
蚀
用sudo运行eclipse
1
2021-09-26 12:42:15
8.27MB
1
3-D convolutional neural networks (3-D-convNets)
have been very recently proposed for action recognition in
videos, and promising results are achieved. However, existing 3-
D-convNets has two “artificial” requirements that may reduce the
quality of video analysis: 1) It requires a fixed-sized (e.g., 112×112)
input video; and 2)most of the 3-D-convNets require a fixed-length
input (i.e., video shots with fixed number of frames). To tackle
these issues, we propose an end-to-end pipeline named Two-stream
3-D-convNet Fusion, which can recognize human actions in videos
of arbitrary size and length using multiple features. Specifically,
we decompose a video into spatial and temporal shots. By taking
a sequence of shots as input, each stream is implemented using
a spatial temporal pyramid pooling (STPP) convNet with a long
short-term memory (LSTM) or CNN-E model, softmax scores of
which are combined by a late fusion.We devise the STPP convNet to
extract equal-dimensional descriptions for each variable-size shot,
andwe adopt theLSTM/CNN-Emodel to learn a global description
for the input video using these time-varying descriptions. With
these advantages, our method should improve all 3-D CNN-based
video analysis methods. We empirically evaluate our method for
action recognition in videos and the experimental results show that
our method outperforms the state-of-the-art methods (both 2-D
and 3-D based) on three standard benchmark datasets (UCF101,
HMDB51 and ACT datasets).
2021-09-25 11:29:08
983KB
1