一些英文写作里的常见实验表述(持续更新中)

最近在写论文,整理了一些英文写作里的常用的表达
领域:视频插帧

  • We summarize the results in 我们总结了…实验结果
  • surpasses all the others in PSNR 在PSNR值上超过了其他方法
  • attribute to the incorporation of 归因于…的利用
  • show visual comparisons of different approaches 展示了不同方法的可视化比较
  • To show the advantage of our framework 为了展示这个框架的优越
  • We suspect that 我们猜测
  • conducted on the GoPro dataset.在某数据集下进行的
  • To justify 为了公正
  • we report the results under several settings .我们报告不同设置下的结果
  • we evaluate the effectiveness of sth concept by replacing some module我们通过替换掉一些模块验证了某理论的有效性
  • we evaluate the contributions of sth 我们验证了…的贡献
  • Excluding some cues 排除掉一些线索
  • we visualize the results of 我们对结果进行可视化
  • renders more accurate transfer with less artifact.有着更精确的转移并且有更少的伪影
  • By default 默认情况下
  • The PSNR of some module degrades much slower than those of others PSNR下降得比其他方法更慢
  • quantitatively and qualitatively compare Net with recent state-of-the-arts on various benchmarks 定性定量的在基准数据集上比较了我们的网络和一些SOTA方法
  • ablation studies are carried out to analyze the contribution of 我们进行了对某模块的贡献的消融实验
  • we execute the released codes of Net and refer to the other results tested in ABME.我们对某网络执行了发布的代码,对其他网络参考了ABME的测试结果
  • For fair comparison 为了公平的比较
  • achieves the best results on both PSNR and SSIM metrics 在PSNR和SSIM指标上都有着最好的效果
  • Our large model also obtains leading results 我们最大的模型取得了最好的结果
  • falls behind Net on the Extreme part. 在极端情况下比我们的网路落
  • attribute the reason to be that将某结果归因于
  • thanks to the superiority of 由于它在某方面的优越性
  • with 15.2 × faster inference speed 有着快乐15.2倍的推理速度
  • reduce parameters and computation complexity by 44% than 比其他降低 了44% 的参数量
  • This can be easily shown 这可以容易的显示
  • Removing the context network results in larger errors 移除掉某个网络会导致大的错误
  • this residual version slightly improves the performance 这个残差版本提升了一点点
  • suggesting that Net may have a more appropriate capacity for this task 更适合这项任务
  • is the fastest among all the top-performing methods 所有排在前面的方法里最快的
  • has about 17 times fewer parameters than 有着17倍小的参数量
  • achieves impressive performance 取得好的效果
  • A larger-capacity optical flow estimator leads to better performance. 大容量光流的效果更好
  • we empirically evaluate the efficiency of some approaches including 我们通过实验验证了方法的高效
  • is discussed in Section 3 在第三章讨论了
  • with only 25% points being retained in each step 在每一步只保留25%的点
  • incur a heavy or limiting computation burden 造成沉重或有限的计算负担
  • Motivated by this, 受此启发
  • To study the importance of additional information provided by events 学习额外事件提供的重要性
  • One version has access to the events 一个版本中有事件
  • We highlight the best performing method. 我们加粗显示了最好的方法
  • outperforms the methods by a margin of 4.24dB. 超过了方法4.24dB
  • efficiently decode the asynchronous intermediate events 有效的解码中间的异步事件
  • As it is clear from the figure图中可以清晰地看到
  • we provide additional details about the training procedure 我们提供了关于训练过程的额外细节
  • get a 0.13 dB improvement 有一个0.13dB的提升
  • prefers synthesis features on coarse scales and warping features on fine scales. 倾向于在粗尺度合成特征,在细尺度光流扭曲
  • facilitate intermediate frame prediction 有助于中间帧的预测
  • our method compares favorably against the state-of-the-art approaches 我们的方法优于先进的方法
  • pose challenges for给…造成了挑战
  • severely blurred videos 严重模糊的视频
  • assume the motion to be linear, which is not always valid in real-world scenarios,假设运动为线性,但是在现实中并不总是成立
  • Recent works have revealed the advantages of sth in area.一些工作揭示了用什么方法在什么领域的好处
  • alleviate motion blur 消除运动模糊
  • inherently embeds 固有的嵌入了
  • compensate the missing information 补偿遗失的信息
  • the problems have strong co-occurrence in real scenes 这两个问题在现实中的强相关的
  • highly dependent on the quality of reference frames 非常依赖参考帧的质量
  • the reference frames are degraded by 参考帧因…质量不高
  • Illustrative examples of 关于…的示例
  • causes performance drop 造成性能的下降
  • reveal the potential of events 揭示了事件的潜力
  • intrinsic camera noises 固有的相机噪声
  • have approached the task.已经接触过了这个任务
  • the cascaded scheme级联方案
  • propagates deblurring error to the interpolation stage, leading to sub-optimal results 把去糊的错误传给了插帧阶段,导致不太好的结果
  • review the physical generation model of events 回顾了事件的物理生成的模型
  • is severely contaminated by noises受噪声严重污染
  • investigate the contribution of the network 研究了某个网络的贡献
  • Prevailing video frame interpolation algorithms流行的插帧算法
  • hindering them from diverse real-time applications 阻碍了他们应用在不同的实时应用上
  • devise a network 设计了一个网络
  • extract pyramid features 提取了金字塔特征
  • fully release its potential 更好的释放它的潜能
  • abrupt brightness change突然的亮度变化
  • suffer from several defects 有以下的缺点
  • increase the inference delay加剧了推理的时延
  • block them from mobile and real-time applications 阻碍了他们在移动实时端的应用
  • merge separated… into a single model 把分开了两个部分融进了一个网络中
  • endowing our model with the ability to do 是我们的模型能去…
  • have their own merits and flaws 有各自的优缺点
  • sketches the overall architecture 描绘了整个架构
  • gradually decimate the spatial size 逐渐下采样
  • target temporal super-resolution on a premise 将时域超分建立在一个前提下
  • experimentally verify this notion 用实验证明了这个概念
  • far less attention has been given to this problem 这个问题关注得太少
  • we conjecture that 我们推测
  • several widely used datasets 一些广泛应用的数据集
  • confirm the favorability of our approach 证实了我们的方法的优势
  • degrade the video quality 降低视频的质量
  • the leftmost subfigure最左边的子图
  • relaxes the constraint of 放宽了…的约束
  • is superior to the second place by 0.08dB 相比第二超过了0.08dB
  • has arisen continuous attention 引起了持续的关注
  • a new upsurge in industry工业界的新热潮
  • The motion field is halved 运动场被减半
  • convey valid motion information between pixel pairs 在像素对中传递有效的运动信息
  • achieves competitive results取得了好的结果
  • override the side effect of resolution loss覆盖分辨率损失的副作用
  • In retrospect of previous work 回顾之前的工作
  • is motion agnostic 是运动不可知的
  • Recent years have witnessed numerous backbone networks 近年许多骨干网络兴起
  • input information follows a fine-to-coarse fashion 输入信息遵循一个从 fine-to-coarse的流程
  • alleviating the gradient vanishing problem 避免梯度消失
  • save the computational budget 节约计算量
  • the number of FLOPs for an image of 224 × 224 pixels is around 4.2G Flops计算
  • We conduct experiments on the ImageNet dataset 我们在某个数据集上进行了实验

你可能感兴趣的:(深度学习)