论文名称:YOLOv4: Optimal Speed and Accuracy of Object Detection

YOLO官网:https://github.com/pjreddie/darknetgithub.com

论文下载:https://arxiv.org/abs/2004.10934

代码下载:https://github.com/AlexeyAB/darknet

作者:Alexey Bochkovskiy, Chien-Yao Wang, Hong-Yuan Mark Liao

发表年份:2020

**地位:**YOLOv4 没有理论创新,而是在原有YOLO目标检测架构的基础上增加了近年CNN改进的众多技术,从数据处理到网络训练再到损失函数,遵行“拿来主义”,加上漂亮的工程实践,打造实现最佳速度与精度平衡的目标检测新基准!

一、 摘要

There are a huge number of features which are said to improve Convolutional Neural Network (CNN) accuracy. Practical testing of combinations of such features on large datasets, and theoretical justifification of the result, is required. Some features operate on certain models exclusively and for certain problems exclusively, or only for small-scale datasets; while some features, such as batch-normalization and residual-connections, are applicable to the majority of models, tasks, and datasets. We assume that such universal features include Weighted-Residual-Connections (WRC), Cross-Stage-Partial-connections (CSP), Cross mini-Batch Normalization (CmBN), Self-adversarial-training (SAT) and Mish-activation. We use new features: WRC, CSP, CmBN, SAT, Mish activation, Mosaic data augmentation, CmBN, DropBlock regularization, and CIoU loss, and combine some of them to achieve state-of-the-art results: 43.5% AP (65.7% AP50*) for the MS COCO dataset at a realtime speed of ∼65 FPS on Tesla V100.*

二、研究背景

2.1 研究现状

总而言之,普通的检测器由以下几个部分组成:

Untitled

2.2 Bag of freebies(指那些能够提高精度而不增加推断时间的技术)

通常,传统的物体检测器是离线训练的。因此,研究人员一直喜欢采用这种优势并开发出更好的训练方法,从而可以使目标检测器获得更好的精度而又不会增加推理成本。我们称这些仅改变培训策略或仅增加培训成本的方法为“Bag of freebies”,这类方法主要有以下: