Baike.dev
All toolsOnline toolsAI codingTrendingOpen sourceNewsSubmit
Log in
< Back to tools
R

R-CenterNet

> 编程语言
Open source

detector for rotated-object based on CenterNet/基于CenterNet的旋转目标检测

400 stars0 likes0 views
WebsiteGitHub

About

detector for rotated-object based on CenterNet/基于CenterNet的旋转目标检测

R-CenterNet(中文)

基于CenterNet的旋转目标检测

前言

本工作初衷是提供一个极其精简的CenterNet代码,并对旋转目标进行检测,1.0为:

  ${R-CenterNet_ROOT}
  |-- backbone
  `-- |-- dlanet.py
      |-- dlanet_dcn.py
  |-- Loss.py
  |-- dataset.py
  |-- train.py
  |-- predict.py

应读者需求,随后更新了2.0

  ${R-CenterNet_ROOT}
  |-- labelGenerator
  `-- |-- Annotations
      |-- voc2coco.py
  |-- evaluation.py

2.0以及data/airplane、imgs、ret文件夹都不是必须的,如果您只是想快速上手,1.0足够了。

demo

  • R-DLADCN(推荐)(DCN编译与原版CenterNet保持一致)
  • R-ResDCN(主干网用的ResNet而不是DLA)
  • R-DLANet(如果你不会编译DCN,就使用这个没有编译DCN的主干网)
  • DLADCN.jpg

常见问题

  • 我对CenterNet原版代码 进行了重构,使代码看起来更加简洁。
  • 如何编译DCN以及环境需求, 与CenterNet 原版保持一致。
  • 关于数据处理与更多细节, 可以参考 here
  • torch版本1.2,如果你用的0.4会发生报错。

训练自己的多分类网络

  • 打标签用labelGenerator文件夹里面的代码。
  • 修改代码中所有num_classes为你的类别数目,并且修改back_bone中hm的数目为你的类别数,如: def DlaNet(num_layers=34, heads = {'hm': your classes num, 'wh': 2, 'ang':1, 'reg': 2}, head_conv=256, plot=False):

Related projects

  • R-CenterNet_onnxruntime【C++】
  • CenterNet

R-CenterNet(English)

detector for rotated-object based on CenterNet

preface

The original intention of this work is to provide a extremely compact code of CenterNet and detect rotating targets: 1.0

  ${R-CenterNet_ROOT}
  |-- backbone
  `-- |-- dlanet.py
      |-- dlanet_dcn.py
  |-- Loss.py
  |-- dataset.py
  |-- train.py
  |-- predict.py

At the request of readers, 2.0 was subsequently updated:2.0

  ${R-CenterNet_ROOT}
  |-- labelGenerator
  `-- |-- Annotations
      |-- voc2coco.py
  |-- evaluation.py

2.0 and the data/airplane, imgs, ret folders are not required. If you just want to get started quickly, 1.0 is enough。

demo

  • R-DLADCN(this code)(How to complie dcn refer to the original code of CenterNet)
  • R-ResDCN(just replace cnn in resnet with dcn)
  • R-DLANet(not use dcn if you don't know how to complie dcn)
  • DLADCN.jpg

notes

  • I refactored the original code to make codes more concise.
  • How to complie dcn and configure the environment, refer to the original code of CenterNet.
  • For data processing and more details, refer to here
  • torch version==1.2,don't use version==0.4!

train your data

  • label your data use labelGenerator;
  • modify all num_classes to your classes num, and modify the num of hm in your back_bone, such as: def DlaNet(num_layers=34, heads = {'hm': your classes num, 'wh': 2, 'ang':1, 'reg': 2}, head_conv=256, plot=False):

Related projects

  • R-CenterNet_onnxruntime【C++】
  • CenterNet

Issues· 21 open

View all issuesOpen on GitHub

No open issues yet, or sync has not completed.

> Tags

Pythoncenternetcomplie-dcndcndetector

No comments yet. Be the first to share.

> Details

PublishedAug 1, 2026
UpdatedSep 17, 2026
Category编程语言
PricingOpen source

> Related tools

T
TypeScript
JavaScript 的超集,为前端与全栈提供静态类型
P
Python
通用编程语言,广泛用于 Web、数据与 AI
G
Go
Google 推出的简洁高效系统语言