1.Markdown语言

Posted 2023-04-09 VincentMTB

Markdown

标题

三级标题

四级标题

字体

Hello, World!

Hello, World!

Hello, World!

Hello, World!

引用

选择JAVA 走向人生巅峰

分割线

---

---

图片

超链接

点击跳转到百度

列表

1. A
2. B
3. C

• A
• B
• C

表格

名字	性别	生日
张三	男	1997.1.1

代码

public

markdown 论文语言素材积累

## Language  
- The main contribuions of our work are summarized as follows: 我们工作的主要贡献如下总结：
- To this end, we present ... 为此，我们提出了...  
- we introduce xxx 我们提出了xxx
- It is simple in formulation，fast and stable in training. 简化建模，训练过程更快更稳定
- we focus on... 我们关注于...
- In computer graphics, two broad approaches to image inpainting exist 在...领域中，存在解决...问题的两大类方法  
- show promising performance in this task. 在这个任务中有很好的效果  
- We achieve an
accuracy of about xx% for the xxx task 我们在xxx任务上取得了约xx的正确率
- our method is inspired by the evolution of recent skeleton-based action recognition using RNN models. 我们的方法是受到最近的RNN模型解决基于骨骼的行为识别方法的启发
- admittedly 公认地，无可否认地
- elaborate 精细的
- for the sake of... 为了 = in order to  
for the sake of accuracy and convebience 为了准确和方便
- owing to  由于  = due to 
- exceedingly rare 极其罕见
- Recently, there have been several attempts to xxx 最近，为了解决xxx有一些尝试
- in the course of... 在...
- should be taken into account 需要被考虑在内
- That is to say 也就是说
- eliminate the effect of xxx 消除xxx造成的影响




## Key points in labanotation and action recognition papers
- motion-captured data 动作捕捉数据
- movement category 运动类别
- movement segments 运动片段
- motion sequence 运动序列（可能是指一系列片段）
- parent joint 父节点 precursor joints（nodes） 前驱节点 
- child joints（nodes） 子节点
- bridge the gap between the motion
capturing technique and Labanotation. 建立运动捕捉和拉班舞谱技术之间的桥梁  
- obtain a robust representation of body movements 获得鲁邦的人体动作表达  
- transform the motion data in the world coordinate system  将运动数据转化到世界坐标系  


**Fusing Geometric Features for Skeleton-Based
Action Recognition using Multilayer LSTM
Networks** 
- considering that LSTM is suitable for modeling dependence in temporal domain 适合对时域依赖性进行建模
- feed xxx network with rich spatial domain features 给网络送入丰富的空间域信息
- explore geometric relations between joints 探索关节之间的几何关系
- adding relations between non-adjacent joints may further enhance the performance. 增加非相邻节点之间的几何关系可能进一步增强效果。
- based on this intuition, we enumerate eight geometric features to describe relations among all joints. 基于这个直觉，我们列举了描述所有节点之间关系的几何特征  
- Compared to the previous xxx, we achieve better performance by integrating all individually trained models with a smoothed score fusion technique. 与先前方法对比，我们通过平滑分数融合技术将多个预训练过的模型相结合，取得了更好的效果
- Vemulapalli et al. utilize rotations and translations to represent the 3D geometric relationships of body parts in Lie group. 在李群中利用旋转平移来描述身体部件之间的3维几何关系
- the fully-connected fusion has almost the same
performance as the single stream 全连接分数融合几乎与单流网络的性能相当
- Since two overfitted networks generate very low loss values, adding
an extra fully connected layer may not improve its overall
performance. 由于两个过拟合网络产生非常低的损耗值，添加额外的全连接层可能不会提高其整体性能


**Automatic Labanotation Generation from Motion-captured Data Based on
Hidden Markov Models**
- Labanotation is a powerful tool for the recording and
archiving of traditional dances. 拉班舞谱是记录和归档传统舞蹈的有力工具
- we extract better features from motion-captured
data that are more conducive to modeling movement segments with Hidden Markov Models. 我们从运动捕捉数据中提取了更好的特征，有助于用HMM模型对运动片段进行建模
- generate much more reliable Labanotation records
than previous works 与之前工作相比可以生成更可靠的拉班舞谱符号
- As a vivid, scientific and logical record system for
human body motion, Labanotation is highly accepted as a
powerful tool for recording human body motion, choreography and dance training    拉班舞谱作为一种生动、科学、逻辑的人体运动记录系统，作为记录人体运动、编舞、舞蹈训练的有力工具，被广泛接受。
- However, writing down the notation by observing the dance movements is a tedious and difficult task which requires patience and skill.  然而，通过观察舞蹈动作来记谱是一项枯燥而困难的任务，需要耐心和技巧。
- Consequently, it is necessary to carry out the research on automatically generating digital Labanotation from real-world dances 因此，有必要对现实舞蹈拉班舞谱自动生成进行研究。
- As an interdisciplinary research area, the technology of automatically generating Labanotation is still in its early stage. The common solution is finding out a technology to acquire human motion data, and then generating Labanotation automatically by computer algorithms.  作为一个跨学科的研究领域，自动生成标记的技术还处于起步阶段。常见的解决方案是找到一种获取人体运动数据的技术，然后用计算机算法自动生成拉班舞谱。
- Motion capturing is the technique of using high dimensional motion data to record the position of the major joints in space, which has been used for measuring human body motion in the fields of CG-animation and movie production. 运动捕捉是利用高维运动数据记录空间主要关节位置的技术，在CG-animation和电影制作领域中已被用于人体运动的测量
- the symbols of support column represent
the transference of weight, which requires dynamic analysis on multiple frames of the motion-captured data across
time.  支撑柱的符号表示重量的传递，需要对运动捕捉数据的多帧进行跨时间的动态分析。
- Optitrack provides a method of documenting movement via recording the movements of 41 reference marker points on the body during a sequence of motion. Optitrack提供了一种通过记录身体上41个参考标记点在运动序列中的运动来记录运动的方法。
- Then the obtained data are integrated into a skeleton model of 26 joints.然后将得到的数据集成到26个关节的骨架模型中
- Finally, the motion-captured data is stored as a BVH (Bio-vision hierarchical) format file. The BVH format of motion-captured data is widely used
for recording human body motions. BVH file saves the rotation data of each joint relative to its parent joint as Euler angle, which depicts the orientation information of joints in the course of human movement. In the processing and analysis of BVH data, the Euler angles are usually converted
to position data in the world coordinate system for convenience. 最后，运动捕捉的数据存储为BVH(生物视觉分级)格式文件。运动捕捉数据的BVH格式被广泛用于记录人体运动。BVH文件将每个关节相对于父关节的旋转数据保存为欧拉角，表示关节在人体运动过程中的方位信息。在BVH数据的处理和分析中，为了方便起见，通常将欧拉角转换为世界坐标系中的位置数据。  
- Suppose that coordinate values of the three joints are (xl; yl; zl),
(xr; yr; zr) and (xs; ys; zs) respectively in the world coordinate system, we can get two intersecting vectors that are determined by the three joints    假设三个关节的坐标值为......在世界坐标系中，我们可以得到由三个关节决定的两个相交向量
- However, different dancers are different in height, shape
and limb lengths. This also leads to a confusion in motion data. We solve this problem by computing the relative position vectors of joints as the feature of normalized motion data. 然而，不同的舞者在身高、体型和肢体长度上是不同的。这也导致了运动数据的混乱。通过计算关节相对位置向量作为归一化运动数据的特征，解决了这一问题。