目录
I. Overview
1. Background
新兴的
The emerging HEVC standard has been developed and standardized collaboratively by both the ITU-T VCEG and ISO/IEC MPEG organizations.
越来越常见
Digital videos are increasingly finding their way into the day-to-day lives of people via the explosion of video applications.
With ConvNets becoming more of a commodity in the computer vision field, ...
越来越受欢迎
Deep CNNs have recently shown an explosive popularity partially due to its success in image classification.
最明显的是\(\cdots\)
There have been significant advances in video processing technology since 2000, most notably the development of the popular AVC compression.
最终用户是人
Large number of applications target humans as the end users of video.
2. Introduction
有以下几个方面
The experimental measurements are mainly of two folds: palaeomagnetism and geochemistry.
轻量级的，简单的
Our deep CNN has a lightweight structure, yet demonstrates state-of-the-art restoration quality.
（旨在）解决\(\cdots\)问题
To address this problem, we propose and extension of ...
We first present a study that we conducted to assess the subjective quality of videos.
A number of attempts have been made to improve the original architecture of Krizhevsky et al. (2012) in a bid to achieve better accuracy.
To this end, we fix other parameters of the architecture, and steadily increase the depth of the network by adding more convolutional layers.
协助，有贡献
We still find that the following local normalization scheme aids generalization.
阻碍
However, the high computational cost still hinders it from practical usage that demands real-time performance (24 fps).
特定领域的
It can be designed to suit domain-specific tasks, i.e., face hallucination.
以\(\cdots\)方式
The explicit enforcement of a photometric distance in the range component of a bilateral filter makes it possible to process color images in a perceptually appropriate fashion.
以\(\cdots\)方式发挥作用
Our method draws its strength from making normalization a part of the model architecture and performing the normalization for each training mini-batch.
考虑到了
Caffe is a deep learning framework made with expression, speed, and modularity in mind.
避免了\(\cdots\)
This eliminates the need to perform most of the SR operation in the far larger HR resolution.
之前提到的
We show that the aforementioned pipeline is equivalent to a deep convolutional neural network.
深入研究
When we delve into the network structure, we find two inherent limitations that restrict its running speed.
补充了\(\cdots\)
This video database presented here, supplements the widely used database for still images.
搭配\(\cdots\)
Current GPUs, paired with a highly-optimized implementation of 2D convolution, are powerful enough.
可行的
To this end, we fix other parameters of the architecture, and steadily increase the depth of the network by adding more convolutional layers, which is feasible due to the use of very small (3 × 3) convolution filters in all layers.
设计
We still find that the following local normalization scheme aids generalization.
直到
We provide a broad view into how to use PyTorch from the basics all the way into constructing deep neural networks.
II. Specification
被称为\(\cdots\)
The only reliable method to assess the video quality is to ask human subjects for their opinion, which is termed subjective video quality assessment (VQA).
病态（欠定）问题
This problem is inherently ill-posed since a multiplicity of solutions exist for any given low-resolution pixel.
It is an underdetermined inverse problem, of which solution is not unique.
（符号）表示，代表着
The observation at any time can be represented by a tensor \(X \in R^{P*M*N}\), where \(R\) denotes the domain of the observed features.
III. Comparison
新的，与众不同的
In this paper, we propose a novel ConvLSTM network for precipitation nowcasting.
比\(\cdots\)更新的（技术，标准）
A number of design aspects new to the HEVC standard improve flexibility for operation over a variety of applications.
一致胜过，当前最好的方法
When evaluated on a synthetic Moving-MNIST dataset and the radar echo dataset, our ConvLSTM model consistently outperforms both the FC-LSTM and the state-of-the-art operational ROVER algorithm.
在某方面胜过
Our model is outperforming that of Szegedy et al. (2014) in terms of the single-network classification accuracy.
在某方面更强的
It is not only a quantitatively superior method, but also a practically useful one.
IV. Number, Magnitude and Extent
大幅度
We can further exceed its accuracy by a substantial margin.
近似，在某个数量级上
Datasets of labeled images were relatively small —— on the order of tens of thousands of images.
快一个数量级
It is an order of magnitude faster than previous CNN-based methods.
占据较大比重
Fiction takes up a large slice of the publishing market.
类似于
Also, analogously to what happens for domain filtering, ...
接近
The current best error rate approaches human performance.
某种程度上说
The parameters used for bilateral filtering in our illustrative examples were to some extent arbitrary.