Technical English (Y2020-P1)

声明：文章内容选自 CVPR 2019：Libra R-CNN: Towards Balanced Learning for Object Detection.

所有内容仅供学习交流使用，不做任何推广。

---

Abstract

这篇文章讲述的是目标检测中的 imbalance 问题，这一问题存在于目标检测的多个环节中。

作者在摘要中这样写道：

Abstract-Sentence 1:

Compared with model architectures, the training process, which is also crucial to the success of detectors, has received relatively less attention in object detection.

为什么要这样写？是因为现在的目标检测算法大多注重在模型框架方面进行改进，但是模型设计出来以后，训练过程难道就是完美的吗？它里边真的就一点问题都没有吗？但是鲜有人对目标检测模型的训练过程进行过系统的研究，尤其是在处理训练过程中出现的 imbalance 问题方面。

于是在指出当前研究中出现的问题之后，作者紧接着对这些问题做了一个更加详细的阐述：

Abstract-Sentence 2:

In this work, we carefully revisit the standard training practice of detectors, and find that the detection performance is often limited by the imbalance during the training process, which generally consists in three levels — sample level, feature level, and objective level.

在这句话中，作者提到，他们重新研究了整个标准的训练过程，从中发现：检测器的性能经常会受到三个层面的 imbalance 问题的影响，即 sample 层面、feature 层面以及 objective 层面。

然后以一种总括的视角引出了自己的工作：

Abstract-Sentence 3:

To mitigate the adverse effects caused thereby, we propose Libra R-CNN, a simple but effective framework towards balanced learning for object detection.

单词：mitigate [vt.] 缓解；减轻

短语：adverse effect(s) 副作用；不利的影响

在这句话中，作者提出，为了缓解在以上三个层面上的 imbalance 问题，他们提出了一个新的框架 —— Libra R-CNN。这一框架的目的是 towards balanced learning for object detection.

Abstract-Sentence 4:

It integrates three novel components: IoU-balanced sampling, balanced feature pyramid, and balanced L1 loss, respectively for reducing the imbalance at sample, feature, and objective level.

作者提出，Libra R-CNN 集成了三个创新的组件（注意这里 integrate 的使用）：IoU-balanced sampling, balanced feature pyramid 以及 balanced L1 loss。这三个组件分别负责在 sample level，feature level 以及 objective level 去解决 imbalance 问题。

Abstract-Sentence 5:

Benefitted from the overall balanced design, Libra R-CNN significantly improves the detection performance.

这里夸了一下 Libra R-CNN。

注意这里的表达方法：significantly improves the detection performance.

Abstract-Sentence 6:

Without bells and whistles, it achieves 2.5 points and 2.0 points higher Average Precision (AP) than FPN Faster R-CNN and RetinaNet respectively on MSCOCO.

这一句是摘要部分的最后一句话，主要是从量化的角度描述了 Libra R-CNN 究竟有多大的改进。

1. Introduction

Introduction-Paragraph 1-Sentence 1:

Along with the advances in deep convolutional networks, recent years have seen remarkable progress in object detection.

短语：along with

表达：recent years have seen …; remarkable progress

这句话是 introduction 部分的总领句。首先从一个非常大的角度去描述了 CNN 的发展对目标检测带来的影响。

Introduction-Paragraph 1-Sentence 2:

A number of detection frameworks such as Faster R-CNN, RetinaNet, and Cascaded R-CNN have been developed, which have substantially pushed forward the state of the art.

单词：develop，这里要表达 “这些框架被提出”，作者用的是 develop 这个单词，而不是 propose。

单词：substantially 本质上；实质上；充分地；相当多地

短语：push forward 奋力向前；推进

表达方式：push forward the state of the art

这句话简单回顾了目标检测中的一些经典框架，然后指出它们曾经产生过很深远的影响。

Introduction-Paragraph 1-Sentence 3:

Despite the apparent differences in the pipeline architectures, e.g. single-stage vs. two-stage, modern detection frameworks mostly follow a common training paradigm, namely, sampling regions, extracting features therefrom, and then jointly recognizing the categories and refining the locations under the guidance of a standard multi-task objective function.

单词：apparent 明显的；显而易见的

单词：paradigm 范例；样式；模范

表达：follow a … paradigm 遵从 … 样的一种规范

表达：under the guidance of

这句话是第一段的最后一句话，作者想表达的意思是：在现在的众多目标检测框架中，它们的结构可能会千差万别，但它们从图片中检测物体的方式却是大同小异的，具体表现在：它们都会经历 region sampling, feature extracting, 以及 jointly classification and localization 的过程。