YOLO全面回顾：从V1到V8资源-CSDN文库

YOLO

需积分: 1 92 浏览量 2024-06-12 07:42:01 上传评论收藏 2.93MB PDF 举报

资源推荐

资源详情

资源评论

A COMPREHENSIVE REVIEW OF YOLO: FROM YOLOV1 TO

YOLOV8 AND BEYOND

UNDER REVIEW IN ACM COMPUTING SURVEYS

Juan R. Terven

CICATA-Qro

Instituto Politecnico Nacional

Mexico

jrtervens@ipn.mx

Diana M. Cordova-Esparaza

Facultad de Informática

Universidad Autónoma de Querétaro

Mexico

diana.cordova@uaq.mx

April 4, 2023

ABSTRACT

YOLO has become a central real-time object detection system for robotics, driverless cars, and video

monitoring applications. We present a comprehensive analysis of YOLO’s evolution, examining the

innovations and contributions in each iteration from the original YOLO to YOLOv8. We start by

describing the standard metrics and postprocessing; then, we discuss the major changes in network

architecture and training tricks for each model. Finally, we summarize the essential lessons from

YOLO’s development and provide a perspective on its future, highlighting potential research directions

to enhance real-time object detection systems.

Keywords YOLO · Object detection · Deep Learning · Computer Vision

1 Introduction

Real-time object detection has emerged as a critical component in numerous applications, spanning various ﬁelds

such as autonomous vehicles, robotics, video surveillance, and augmented reality. Among the various object detection

algorithms, the YOLO (You Only Look Once) framework has stood out for its remarkable balance of speed and accuracy,

enabling the rapid and reliable identiﬁcation of objects in images. Since its inception, the YOLO family has evolved

through multiple iterations, each building upon the previous versions to address limitations and enhance performance

(see Figure 1). This paper aims to provide a comprehensive review of the YOLO framework’s development, from the

original YOLOv1 to the latest YOLOv8, elucidating the key innovations, differences, and improvements across each

version.

The paper begins by exploring the foundational concepts and architecture of the original YOLO model, which set the

stage for the subsequent advances in the YOLO family. Following this, we delve into the reﬁnements and enhancements

introduced in each version, ranging from YOLOv2 to YOLOv8. These improvements encompass various aspects such

as network design, loss function modiﬁcations, anchor box adaptations, and input resolution scaling. By examining

these developments, we aim to offer a holistic understanding of the YOLO framework’s evolution and its implications

for object detection.

In addition to discussing the speciﬁc advancements of each YOLO version, the paper highlights the trade-offs between

speed and accuracy that have emerged throughout the framework’s development. This underscores the importance of

considering the context and requirements of speciﬁc applications when selecting the most appropriate YOLO model.

Finally, we envision the future directions of the YOLO framework, touching upon potential avenues for further research

and development that will shape the ongoing progress of real-time object detection systems.

UNDER REVIEW IN ACM COMPUTING SURVEYS

2021

YOLOv1

YOLO9000

YOLOv3

Scaled

YOLOv4

PP-YOLO

YOLOv5

YOLOv6

YOLOX

YOLOR

PP-YOLOv2

DAMO YOLO

PP-YOLOE

YOLOv7

YOLOv6

2015

2016

2018

2020

2022

2023

YOLOv8

Figure 1: A timeline of YOLO versions.

2 YOLO Applications Across Diverse Fields

YOLO’s real-time object detection capabilities have been invaluable in autonomous vehicle systems, enabling quick

identiﬁcation and tracking of various objects such as vehicles, pedestrians [

], bicycles, and other obstacles [

These capabilities have been applied in numerous ﬁelds, including action recognition [

] in video sequences for

surveillance [8], sports analysis [9], and human-computer interaction [10].

YOLO models have been used in agriculture to detect and classify crops [

], pests, and diseases [

], assisting in

precision agriculture techniques and automating farming processes. They have also been adapted for face detection

tasks in biometrics, security, and facial recognition systems [14, 15].

In the medical ﬁeld, YOLO has been employed for cancer detection [

], skin segmentation [

], and pill

identiﬁcation [

], leading to improved diagnostic accuracy and more efﬁcient treatment processes. In remote sensing,

it has been used for object detection and classiﬁcation in satellite and aerial imagery, aiding in land use mapping, urban

planning, and environmental monitoring [20, 21, 22, 23].

Security systems have integrated YOLO models for real-time monitoring and analysis of video feeds, allowing rapid

detection of suspicious activities [

], social distancing, and face mask detection [

]. The models have also been

applied in surface inspection to detect defects and anomalies, enhancing quality control in manufacturing and production

processes [26, 27, 28].

In trafﬁc applications, YOLO models have been utilized for tasks such as license plate detection [

] and trafﬁc

sign recognition [

], contributing to the development of intelligent transportation systems and trafﬁc management

solutions. They have been employed in wildlife detection and monitoring to identify endangered species for biodiversity

conservation and ecosystem management [

]. Lastly, YOLO has been widely used in robotic applications [

] and

object detection from drones [34, 35].

3 Object Detection Metrics and Non-Maximum Suppression (NMS)

The Average Precision (AP), traditionally called Mean Average Precision (mAP), is the commonly used metric for

evaluating the performance of object detection models. It measures the average precision across all categories, providing

a single value to compare different models. The COCO dataset makes no distinction between AP and AP. In the rest of

this paper, we will refer to this metric as AP.

In YOLOv1 and YOLOv2, the dataset utilized for training and benchmarking was PASCAL VOC 2007, and VOC 2012

[

]. However, from YOLOv3 onwards, the dataset used is Microsoft COCO (Common Objects in Context) [

]. The

AP is calculated differently for these datasets. The following sections will discuss the rationale behind AP and explain

how it is computed.

UNDER REVIEW IN ACM COMPUTING SURVEYS

3.1 How AP works?

The AP metric is based on precision-recall metrics, handling multiple object categories, and deﬁning a positive

prediction using Intersection over Union (IoU).

Precision and Recall

: Precision measures the accuracy of the model’s positive predictions, while recall measures the

proportion of actual positive cases that the model correctly identiﬁes. There is often a trade-off between precision and

recall; for example, increasing the number of detected objects (higher recall) can result in more false positives (lower

precision). To account for this trade-off, the AP metric incorporates the precision-recall curve that plots precision

against recall for different conﬁdence thresholds. This metric provides a balanced assessment of precision and recall by

considering the area under the precision-recall curve.

Handling multiple object categories

: Object detection models must identify and localize multiple object categories

in an image. The AP metric addresses this by calculating each category’s average precision (AP) separately and then

taking the mean of these APs across all categories (that is why it is also called mean average precision). This approach

ensures that the model’s performance is evaluated for each category individually, providing a more comprehensive

assessment of the model’s overall performance.

Intersection over Union

: Object detection aims to accurately localize objects in images by predicting bounding

boxes. The AP metric incorporates the Intersection over Union (IoU) measure to assess the quality of the predicted

bounding boxes. IoU is the ratio of the intersection area to the union area of the predicted bounding box and the ground

truth bounding box (see Figure 2). It measures the overlap between the ground truth and predicted bounding boxes.

The COCO benchmark considers multiple IoU thresholds to evaluate the model’s performance at different levels of

localization accuracy.

Figure 2: Intersection over Union (IoU). a) The IoU is calculated by dividing the intersection of the two boxes by the

union of the boxes; b) examples of three different IoU values for different box locations.

3.2 Computing AP

The AP is computed differently in the VOC and in the COCO datasets. In this section, we describe how it is computed

on each dataset.

VOC Dataset

This dataset includes 20 object categories. To compute the AP in VOC, we follow the next steps:

For each category, calculate the precision-recall curve by varying the conﬁdence threshold of the model’s

predictions.

UNDER REVIEW IN ACM COMPUTING SURVEYS

Calculate each category’s average precision (AP) using an interpolated 11-point sampling of the precision-recall

curve.

3. Compute the ﬁnal average precision (AP) by taking the mean of the APs across all 20 categories.

Microsoft COCO Dataset

This dataset includes 80 object categories and uses a more complex method for calculating AP. Instead of using an

11-point interpolation, it uses a 101-point interpolation, i.e., it computes the precision for 101 recall thresholds from 0

to 1 in increments of 0.01. Also, the AP is obtained by averaging over multiple IoU values instead of just one, except

for a common AP metric called

, which is the AP for a single IoU threshold of 0.5. The steps for computing AP in

COCO are the following:

For each category, calculate the precision-recall curve by varying the conﬁdence threshold of the model’s

predictions.

2. Compute each category’s average precision (AP) using 101-recall thresholds.

Calculate AP at different Intersection over Union (IoU) thresholds, typically from 0.5 to 0.95 with a step size

of 0.05. A higher IoU threshold requires a more accurate prediction to be considered a true positive.

4. For each IoU threshold, take the mean of the APs across all 80 categories.

5. Finally, compute the overall AP by averaging the AP values calculated at each IoU threshold.

The differences in AP calculation make it hard to directly compare the performance of object detection models across

the two datasets. The current standard uses the COCO AP due to its more ﬁne-grained evaluation of how well a model

performs at different IoU thresholds.

3.3 Non-Maximum Suppression (NMS)

Non-Maximum Suppression (NMS) is a post-processing technique used in object detection algorithms to reduce the

number of overlapping bounding boxes and improve the overall detection quality. Object detection algorithms typically

generate multiple bounding boxes around the same object with different conﬁdence scores. NMS ﬁlters out redundant

and irrelevant bounding boxes, keeping only the most accurate ones. Algorithm 1 describes the procedure. Figure 3

shows the typical output of an object detection model containing multiple overlapping bounding boxes and the output

after NMS.

Algorithm 1 Non-Maximum Suppression Algorithm

Require: Set of predicted bounding boxes B, conﬁdence scores S, IoU threshold τ, conﬁdence threshold T

Ensure: Set of ﬁltered bounding boxes F

1: F ← ∅

2: Filter the boxes: B ← {b ∈ B | S(b) ≥ T }

3: Sort the boxes B by their conﬁdence scores in descending order

4: while B 6= ∅ do

5: Select the box b with the highest conﬁdence score

6: Add b to the set of ﬁnal boxes F : F ← F ∪ {b}

7: Remove b from the set of boxes B: B ← B − {b}

8: for all remaining boxes r in B do

9: Calculate the IoU between b and r: iou ← IoU(b, r)

10: if iou ≥ τ then

11: Remove r from the set of boxes B: B ← B − {r}

12: end if

13: end for

14: end while

We are ready to start describing the different YOLO models.

4 YOLO: You Only Look Once

YOLO by Joseph Redmon et al. was published in CVPR 2016 [

]. It presented for the ﬁrst time a real-time end-to-end

approach for object detection. The name YOLO stands for "You Only Look Once," referring to the fact that it was

UNDER REVIEW IN ACM COMPUTING SURVEYS

Figure 3: Non-Maximum Suppression (NMS). a) Shows the typical output of an object detection model containing

multiple overlapping boxes. b) Shows the output after NMS.

able to accomplish the detection task with a single pass of the network, as opposed to previous approaches that either

used sliding windows followed by a classiﬁer that needed to run hundreds or thousands of times per image or the more

advanced methods that divided the task into two-steps, where the ﬁrst step detects possible regions with objects or

regions proposals and the second step run a classiﬁer on the proposals. Also, YOLO used a more straightforward output

based only on regression to predict the detection outputs as opposed to Fast R-CNN [

] that used two separate outputs,

a classiﬁcation for the probabilities and a regression for the boxes coordinates.

4.1 How YOLOv1 works?

YOLOv1 uniﬁed the object detection steps by detecting all the bounding boxes simultaneously. To accomplish

this, YOLO divides the input image into a

S × S

grid and predicts

bounding boxes of the same class, along

with its conﬁdence for

different classes per grid element. Each bounding box prediction consists of ﬁve values:

P c, bx, by, bh, bw

where

P c

is the conﬁdence score for the box that reﬂects how conﬁdent the model is that the box

contains an object and how accurate the box is. The

and

coordinates are the centers of the box relative to the grid

cell, and

and

are the height and width of the box relative to the full image. The output of YOLO is a tensor of

S × S × (B × 5 + C) optionally followed by non-maximum suppression (NMS) to remove duplicate detections.

In the original YOLO paper, the authors used the PASCAL VOC dataset [

] that contains 20 classes (

C = 20

); a grid

of 7 × 7 (S = 7) and at most 2 classes per grid element (B = 2), giving a 7 × 7 × 30 output prediction.

Figure 4 shows a simpliﬁed output vector considering a three-by-three grid, three classes, and a single class per grid for

eight values. In this simpliﬁed case, the output of YOLO would be 3 × 3 × 8.

YOLOv1 achieved an average precision (AP) of 63.4 on the PASCAL VOC2007 dataset.

4.2 YOLOv1 Architecture

YOLOv1 architecture comprises 24 convolutional layers followed by two fully-connected layers that predict the

bounding box coordinates and probabilities. All layers used leaky rectiﬁed linear unit activations [

] except for the

last one that used a linear activation function. Inspired by GoogLeNet [

] and Network in Network [

], YOLO uses

1 × 1

convolutional layers to reduce the number of feature maps and keep the number of parameters relatively low. As

activation layers, Table 1 describes the YOLOv1 architecture. The authors also introduced a lighter model called Fast

YOLO, composed of nine convolutional layers.

4.3 YOLOv1 Training

The authors pre-trained the ﬁrst 20 layers of YOLO at a resolution of

224 × 224

using the ImageNet dataset [

]. Then,

they added the last four layers with randomly initialized weights and ﬁne-tuned the model with the PASCAL VOC 2007,

and VOC 2012 datasets [36] at a resolution of 448 × 448 to increase the details for more accurate object detection.

For augmentations, the authors used random scaling and translations of at most 20% of the input image size, as well as

random exposure and saturation with an upper-end factor of 1.5 in the HSV color space.

剩余26页未读，继续阅读

评论收藏

内容反馈

jluliuchao

粉丝: 6
资源: 241

YOLO全面回顾：从V1到V8

中文翻译学习笔记-YOLO的全面评述：从YOLOv1到YOLOv8

YOLO核心思想：从R-CNN到Fast R-CNN一直采用的思路是proposal+分类 （proposal 提供位置信息，

为YOLO V5铺垫：一文看懂YOLO V1-V4的变化

从YOLOV1到V8的YOLO全面回顾

【YOLO开发实战】：从入门到精通的深度学习之旅-Markdown文章材料.zip

yolo论文系列-v1-v8

YOLO开发教程：从零开始构建自己的目标检测系统.md

yolotfjs-yolo、yolo v3和Tiny yolo v1、v2、v3与Tensorflow.js

目标检测经典论文-YOLO论文翻译：（YOLO：统一的实时目标检测）

yolo-labeler：每个图像针对单个对象的自动YOLO标签

Yolo-PyTorch:YOLO v1在PyTorch中的实现

YOLO_tensorflow:YOLO的tensorflow实现

YOLO-Former：YOLO与ViT握手

YOLO-TLA：基于YOLOv5的高效轻量级小目标检测模型

通俗易懂的YOLO系列（从V1到V5）模型解读！

YOLO-World：实时开放词汇对象检测

YOLO 数据集：匕首、刀片检测（1类别，包含训练集、验证集）

YOLO9000: Better, Faster, Stronger 论文

YOLO-Drone：高空视角空中实时检测致密小物体

OpenCV 4.8.0

ZXPSignLib-minimal.dll

落雪音乐自定义音源切换

OriginPro 色卡

基于FPGA的ov5640图像采集

Video Speed Controller谷歌插件

GIF图片制作神器gifcam.exe

python调用DXGI实时快速截屏，是python截屏的最快版了

DisplayPort (DP) 协议标准 V1.4-中文版

xfeatures2d.zip

最新资源

YOLO核心思想：从R-CNN到Fast R-CNN一直采用的思路是proposal+分类（proposal 提供位置信息，