Introduction
Introduction的意义。出处忘记了:sweat_smile:，与大家共勉。
introduction重点在于说明自己的研究思路，很多人搞了半辈子的研究，到头来写文章时忘记的研究的初衷。好的套路就是把读者当成你的同行，讲清楚研究工作的来龙去脉。下面分三步说明。第一步，你发现有什么问题，这篇论文要解决什么问题？实际工作中存在什么问题。不要扯太远，尽量不要超过3句话。这第一步可能就把很多人难住了。尼玛课题都是老板定的，有个毛意义呀，不知道，就瞎扯一个。如果实在不清楚，说明占有的文献还不够，还得多读读。第二步，基于本人前期的研究基础，或者受前人的思路激发，我想出了方法来解决第一步的问题。这一步是关键。很多人的introduction根本没这些内容。完全就是一副“这玩意以前没人搞过，我用一种方法来试试”的态度。这一步为什么非常关键？因为同样的问题，如果很重要，很有意义的问题，肯定很多人想到了，为什么别人没做，而你想做？只有2种可能，我这里有研究基础，别人做不了，我能做。还有一种可能，这里面有难点，我也是受哪篇哪篇文章的激发，才想起解决方法，所以我才顺便试试，结果你看，一搞，结果挺好，所以来跟你们报告一下。这里面就需要吹吹牛逼了，可以弹压一下前人的工作，鼓吹一下自己的工作。第三步，前面牛吹了，后面当然要简单介绍一下结果，谈点意义，勾引读者往下看研究方法。论文的重点在于论，要讲清楚研究思路，而不是搞成实验报告。
Content
展示的内容是按四段式分，当然每个部分也可以拆成几个小段，不要拘泥于此。提供了多个样例，灵活组合吧。
第一段
引出你研究工作的背景和价值
| 一个大红大热的方向（吸引了很多关注），主要思路是怎么做呢，有什么优点 |
|---|
| An explosively growing line of research on xx studies a new paradigm of yy: [details of the paradigm] |
| xx has enjoyed tremendous growth and exhibited a wealth of development at both the conceptual and application levels |
| In recent years, with the vigorous development of xx yy has attracted more and more interest in both zz |
| Successful applications have been demonstrated in areas spanning xx, xx |
| xx has found numerous applications in xx field, particularly in the last two decade |
| a theoretical understanding of GANs is an on-going research topic |
| Abundant research has been conducted on xx |
| 描述方法涵盖了很多方面，子问题 a widely-used technique for solving many kinds of vision and graphics tasks ranging from view synthesis, to re-lighting, to pose and shape estimation, to 3D-aware image synthesis and editing. |
| xx have become prevalent in recent years. 盛行的，普遍的 |
| 直接从技术名称开始，介绍它的优点｜应用价值，然后第二段再转到它的缺点 |
|---|
| GANs excel at synthesizing photo-realistic images. Their conditional extension, conditional GANs, allows controllable image synthesis and enables many computer vision and graphics applications such xxx |
| 交代你所解决的任务 |
|---|
| We consider the task of ... |
| A key challenge in this task is ... |
| xx addresses this issue by integrating xx into xx and enabling xx. |
第二段
阐述存在的问题（也是你要研究的），然后现有方法是怎么来做的
| 说现有方法的不好之处 |
|---|
| The main drawback of such models |
| lagged far behind 现在新领域的方法远远落后于过去某一领域的技术 |
| A body of works 一批工作 / Prior attempts .. |
| Several follow-up studies propose to ... |
| xx hamper the applicability of the method. （妨碍了他的广泛应用） |
| A prominent difficulty |
| unmet needs of 不满足的需求 |
| One prominent hurdle (突出的障碍) of the existing method is xx |
| xx has a long history [ref 1,2,3] |
| We argue that it is more desirable to xx |
| an inevitable problem 不可避免的问题 has two severe difficulties: |
| The image quality still lags far behind (落后) other methods |
| in contrast, previous work |
| several attempts are being made to ... 现有的一些工作已经在怎么做了 |
| xx is still lagging 现有方法还依旧是落后的 |
| 现存方法大致可以分为几类 |
|---|
| Existing approaches to address this problem can be roughly divided into three categories. The first category of works utilizes xx. The second line of works adopt xx. |
| in contrast, previous work ... / In parallel work |
| A thread of recent research/studies |
| 虽然这些方法怎么怎么样了，但还是没能好好解 决 |
|---|
| While these methods compensate .. by .., this task remains challenging. |
| xx is an important but less explored problem |
| This serves as a major bottleneck, as multiple-view data is hard to acquire. |
| xx circumvent a fundamental difficulty 避开了一个很难的问题 |
| To circumvent xx 别人的工作为了回避 |
| counteract xx 为了抵消什么缺点 |
| 发展很久了但应用到什么领域还是新的 | 这个问题还没有被深入研究，同时很具有研究价值 |
|---|
| Despite its potential to serve as a powerful method to xx, this problem remains much less explored. |
| Despite its practical importance, up to our knowledge, we are the first to address this unexplored problem. |
| Existing approaches can not well | have not been well studied |
| To our knowledge, this is the first study in which the has not been proposed yet |
| 最近开始流行一个新技术 |
|---|
| Looking at the research in recent years, there is an overwhelming evidence in support of xx credited to xx. |
| With the rapid developments in xx |
第三段
讲自己的方法；先笼统地说一下方法，然后具体地 specifically
| 承上启下引入 |
|---|
| To address the above problem, we propose xx |
| To answer this question, we conduct a comprehensive empirical study by xx |
这篇文章提出了|做了什么
一般开头都是 In this work, ….
| 用对比的形式，我们先不是在做什么，而是要做什么 |
|---|
| Instead of investigating xx, we seek to xx. |
| 主动形式｜直截了当 |
|---|
| In this work, we propose a new approach, [Name] |
| We propose to exploit |
| Technically/Concretely/Crucially/Specifically… , we propose a xx model .. |
| In this work we demonstrate that |
| To tackle the problem, we extend |
| Here we put forward a new view of |
| 可以用到的词汇 |
|---|
| To alleviate these issues, we design techniques |
| To compensate for xx 为了缓解 |
| To overcome aforementioned limitations we present a new |
| bridges the gap between xx and yy |
| To tackle the problem, we extend |
| the crux of xxx 关键点 |
| We mitigate problem |
| This model has several benefits, |
| in addition to achieving more compelling and complicated xx, |
最后一段
说本文价值，夸就完事了
| 本文价值，可以是分析了多种方法，比较了好坏 |
|---|
| We analyse and investigate multiple design choices for representing neural implicit surfaces, ranging form monolithic MLP models over single-grid to multi-resolution grid representations. |
| The experiments in the paper would be more compelling |
| 强调虽然本文中我们只用在了某个数据集上，但这个方法可以适用于很多领域 while this study presented in this paper focus on xx, is also applicable to other domains |
| As a result, our approach inherits xx 好处 and yy 好处, seperated from zz 不好的方法, which 具体不好的点。 |
| 在很多数据集上进行了测试 |
|---|
| We empirically demonstrate the advantage of xx over existing approaches on a variety of xx tasks: |
| We evaluate our method on both tasks of 2D image synthesis and 3D-aware image synthesis. On a wide range of datasets including [xx data], [model name] exhibits consistent improvements over the baselines. |
| Qualitative and quantitative results show the superiority of our approach over existing methods in preserving identity as well as generating realistic high quality images. |
| We validate our approach through quantitative and qualitative results and demonstrate that our method achieves state-of-the-art results for the task of StyleGAN inversion and real image editing. |
| Our experiments show that the novel method outperforms previous work |
结尾
| 给出贡献列表 |
|---|
| In summary, our main technical contributions are |
| Overall, the main contributions and benefits of our algorithm are summarized as follows: |
| Our proposed framework has several theoretical and practical contributions: |
| To summarize, our work provides the following novelties: |
| 有关公开代码 |
|---|
| Code and models will be provided upon acceptance. |
| 论文结构 |
|---|
| The reminder of this paper is organized as follows. In Sect. 2, we review previous studies that are closely related to the general scope of our work. In Sect. 3, we introduce specific techniques involved in . In Sect. 4, we discuss different possibilities for . We document the implementation details, experimental settings, and comparative results in Sect. 5. In the end, we conclude this paper in Sect. 7. |
点睛之笔
| 可以用一些疑问句来引出论文的核心 |
|---|
| Our work aims to answer the following question / Can this been done in an unsupervised manner |
| The research question we are addressing in this work is, whether an event stream from a DVS moving around the scene is sufficient to reconstruct a 3D representation of a static scene, e.g., NeRF. |
| What extent has this inability of convolution persisted insidiously inside other tasks |
| Is it possible to reconstruct the 3D shape of a single 2D image by exploiting the 3D-alike image manipulation effects produced by GANs? |
| It naturally raises a question: does a discriminator with a fixed capacity meet the demand of such a dynamic training environment? |
Thus, in this paper, we aim toaddress the following questions: Do MAML and ANIL provably learn high-qualityrepresentations? If so, why? |
Hence, anatural question to ask is, |
| 以及多提一提论文的核心 |
|---|
| In this paper, we focus on xx |
| 给出自己方法的定位 |
|---|
| Our work can be classified as a category-specific 3D aware neural novel view synthesis method. |
| Our [model name] can be categorized into representation learning as well as architecture design for GANs. |
| Closed to our approach is xx, where they xx. Different from their work, we use |
如果是和别人工作的嫁接，组合式的论文
xx marries (y method) with [z method]
our framework combines the benefits of xx with those of yy, enabling zz
we incorporate xx into our model
its effective integration with xx
| 方法不是原创的，但我拿来用到了一个新的领域 |
|---|
| Tuning the generator to accommodate a given input image (描述你的方法) is not new~\cite{}, we first introduce this approach to the field of StyleGAN-based editing. |
| (最好还要写出我不是生硬地套过来，是因地制宜的，或者效果更好，达到了我的目的) In contrast to the aforementioned work, our technique maintains the unique editing capabilities, while preserving identity accurately. |
有时候为了节省篇幅，会省略一些东西，就需要加以说明
| 省略 |
|---|
| for ease of notation, we will simply write the xxx |
| For convenience, we follow the convention for xx by letting subscripts denote absolute yy |
| For clarity, we omit z from hereon forward |
| we dropped the vector arrows for ease of notation. |
| for more details, we refer readers to sb |
| we assume a fixed, and remove the subscript xx notation to simplify notation. |
| which we will call a “diffusion model” for brevity |
| Detailed theoretical analysis will be elaborated later. |