comicom / yolov4

YOLOv4：Optimal Speed and Accuracy of Object Detection

https://hoya012.github.io/blog/yolov4/

https://medium.com/visionwizard

https://github.com/AlexeyAB/darknet

abstract

논문 링크

• You Only Look Once: Unified, Real-Time Object Detection
• YOLO9000: Better, Faster, Stronger
• YOLOv3: An Incremental Improvement
• YOLOv4: Optimal Speed and Accuracy of Object Detection

해설 영상 링크

• You Only Look Once: Unified, Real-Time Object Detection
• YOLO9000: Better, Faster, Stronger
• YOLOv3: An Incremental Improvement
• YOLOv4: Optimal Speed and Accuracy of Object Detection

필요조건

• Tensorflow 2.1.0
• tensorflow_addons 0.9.1 (required for mish activation)

how to install

설치방법

download

yolov4 tutorial

<< best requirment >>

• (darkNet)[https://github.com/AlexeyAB/darknet]
• CUDA 10.0
• cuDNN 7.4.3
• visual studio 2015
• opencv 4.1.0

weight file

yolov4 tutorial

1. Downloading DarkNet

• (Alexeyab darknet)[https://github.com/AlexeyAB/darknet]

2. Copying Open files into DarkNet

3. Copying CuDNN files into DarkNet

4. Changing the CuDNN version in DarkNet

5. Compile YOLOv4 with update CUDA version

6. compiling DarkNet

7. Run Detction on Images

8. Detection on Videos

https://blog.naver.com/shuna11/221979528693

problem1

include / library 디렉토리 잡는 법

프로젝트 > 속성 > VC++ directories include directories

• (cuda 설치 디렉토리)\include
• (opencv\빌드\인스톨)\include

library directories

• (cuda 설치 디렉토리)\lib\x64
• (opencv\빌드\인스톨)\x64\vc14\lib

problem2

CUDA10.0.props & CUDA10.0.targets 잡는 법

line56

<Import Project="(cuda 설치 디렉토리)\extras\visual_studio_integration\MSBuildExtensions\CUDA10.0.props"/>

line308

<Import Project="(cuda 설치 디렉토리)\extras\visual_studio_integration\MSBuildExtensions\CUDA10.0.targets"/>