experiment.md · 浙江理工大学-冯杰/EnglishPaperWriting

# 实验部分（杨小利 ）

The first setting aims to remove the undesired foreground object, which has been explored in the previous work。
> 第一个设置的目的是消除以前的工作中已经探索过的不希望出现的.....

This setting corresponds to some real-world applications such as......
> 此设置对应于一些实际应用，如......

Unless otherwise indicated,we fix.....as......
> 除非另有说明，我们将......设为......

To demonstrate the effectiveness and generalization ability of the ......,we evaluate our method on ....datasets。
> 为了证明该方法的有效性......我们在......数据集上评估了我们的方法。

To prepare the training set, we synthesize and overlay a mask of random shape onto each frame of a video。
> 为了准备训练集，我们在视频的每一帧上合成并覆盖一个随机形状的掩膜。

PSNR measures image’s distortion, while SSIM measures the similarity in structure between the two images
>PSNR测量图像的失真，而SSIM则测量两幅图像的结构相似性。

We quantitatively and qualitatively compare our approach with other existing methods on ..... datasets.
>我们在.....数据集上将我们的方法的其他现有方法进行了数量和质量上的比较。

Our approach achieves the best performance on both datasets.
>我们的方法在这两个数据集上都取得了最佳的性能。

Compared with conventional video inpainting approaches , our approach could better handle videos with complex motions。
>与传统的视频修复方法相比，我们的方法可以更好地处理复杂运动的视频

Meanwhile, our approach is significantly faster in runtime speed and thus it is more well-suited for real-world applications.
>同时，我们的方法在运行时速度明显更快，因此更适合于现实世界的应用。

Our approach significantly outperforms the other two baselines, while the image inpainting method performs the worst since it is not designed to maintain temporal consistency on its output.
>我们的方法明显优于其他两个基线，而图像修复方法的性能最差，因为它的设计并不是为了保持输出的时间一致性。

Figure 6 shows some examples of our inpainting results
>图6显示了一些修复结果的示例

Table 2 presents the inpainting results on.....
>表2显示了在......修复的结果

As introduced in Sec. 3.2，
>正如在3.2节所介绍的。

Similarly, in this experiment, we adopt the ...... to examine the effectiveness of hard flow example mining.
>同样，在这个实验中，我们采用第一个.......来检验硬流示例挖掘的有效性。

Table 3 lists the flow completion accuracy under different mining settings, as well as the corresponding inpainting
performance
>表3列出了在不同挖掘设置下的流完成精度以及相应的修复性能。

For clear demonstration
>为了清楚的展示

The result of ....... is somewhat inferior to the second stage, which suggests the effectiveness of using the stacked architecture in this task.
>......的结果略逊于第二阶段，这表明了在此任务中使用堆叠架构的有效性。

To further indicate the effectiveness of using......,we compare ...... with......
>为了进一步说明.......的有效性，我们将我们的.....和......进行了比较。

This step produces high-quality results and also reduces the size of missing regions that have to be handled in the subsequent step
>此步骤产生高质量的结果，并减少在后续步骤中必须处理的缺失区域的大小。

As shown in Table ....., compared with a baseline approach that ........
>如表.....所示，与.....基准方法相比，

this step greatly eases the task and improves the overall performance
>这个步骤大大简化了任务，提高了总体性能

We initialize the weights using the initialization method described in ..... and use ..... for optimization
> 我们使用.....中描述的初始化方法初始化权重，并使用......进行优化。

Our method is denoted as .....
>我们的方法表示为

Figure 5 and Figure 6 shows the comparisons on ...... and ..... respectively
>图5和图6分别显示了在.....和.....上的比较。

It can be seen that our method outperforms all the other methods on these measurements on.....
>可以看出，我们的方法在这些.....度量方面优于所有其他方法。

In addition to quantitative comparisons, we also evaluate our algorithm via a human subjective study
>除了定量比较之外，我们还通过人的主观研究来评估我们的算法。

In videos, the occluded or removed parts in a frame are often revealed in the past/future frames as the objects move and
the viewpoint changes.
>在视频中，随着物体的移动和视点的变化，在过去/未来的帧中，被遮挡或被移除的部分往往会被显示出来。

Figure 2 depicts the pipeline of our ....... approach.
>图2描绘了我们的......方法的流程图。

It contains two steps, the first step is to ..... while the second step is to ......
>它包含两个步骤,第一步是完成.....,而第二步是......

feeds them into the ...... for .......
>把它们输入......以便.......

Section 3.1 will introduce our basic ....... in detail.
>第3.1节将详细介绍我们的.....

Figure ..... depicts the architecture of.....
>图.....描述了.....的体系结构

After......，the second ....... focuses on further ........
>.....之后，第二......集中于进一步的.......

A step-by-step visualization is provided in Fig. 3
>图3提供了一步的可视化。

For the joint fine-tuning, the overall loss is a linear combination of ...... losses
>对于联合微调，总损耗是......损耗的线性组合。

In our experiments, we observe that directly using ...... generally leads to the ..... problem
>在我们的实验中，我们观察到直接使用......通常会导致.....问题。

What is worse, the ....... can lead to ...... in the ...... step
>更糟糕的是，......在....步骤中会导致.......

To overcome this issue, inspired by ......, we leverage.....
>为了克服这个问题，在.....的启发下，我们利用......

Thanks to the ......, the model learns to focus on producing sharper boundaries.
>由于......，模型学会了专注于产生更清晰的边界。

We download 100 videos of different scenes collected from ......
> 我们从......下载100个不同场景的视频

91 videos (about 40,000 frames) are used as the training dataset and 9 videos are used as the validation dataset
>训练数据集使用91段视频(约40 000帧)，验证数据集使用9段视频。

> All video frames are resized to 640 × 360
所有视频帧都被调整为640×360

浙江理工大学-冯杰/EnglishPaperWriting

简介

发行版

贡献者

近期动态

浙江理工大学-冯杰/EnglishPaperWriting .gitee-modal { width: 500px !important; }

简介

发行版

贡献者

近期动态

搜索帮助

浙江理工大学-冯杰/EnglishPaperWriting