Light

songwsx / rfsong-7993 Goto Github PK

View Code? Open in Web Editor NEW

174.0 9.0 50.0 12.03 MB

设计的轻量级RFB进行行人检测，AP达到0.7993，参数量仅有3.1MB，200 FPS

Python 94.78% C++ 0.14% Cuda 4.97% Shell 0.10%

rfsong-7993's Introduction

RFSong-7993进行行人检测

重新设计的RFBNet300,模型参数量只有0.99MB，AP达到0.80(相比RFBNet300降了两个点)，速度却可以达到200FPS，并且在多个其他任务表现良好，例如钢筋检测，人手检测等等

文章链接：https://zhuanlan.zhihu.com/p/76491446

环境

在python3.6 Ubuntu16.04 pytorch1.1下进行了实验

编译NMS

./make.sh

训练自己的数据集

新建一个文件夹VOCdevkit，将自己的VOC格式数据集的VOC2007文件夹移动到VOCdevkit中，例如我的就是/home/common/wangsong/VOCdevkit/VOC2007

$VOCdevkit/
$VOCdevkit/VOC2007/
$VOCdevkit/VOC2007/Annotations/
$VOCdevkit/VOC2007/ImageSets/
$VOCdevkit/VOC2007/JPEGImages/

修改data/config.py中的数据集路径 VOCroot = '/home/common/wangsong/VOC/VOCdevkit' 改为自己数据集对应的路径
修改针对VOC行人的代码部分，可以参考我的博客查看修改的地方

train_RFB.py的trainsets进行改动：train_sets = [('2007', 'trainval'))]
data/voc0712.py和data/voc_eval.py这两个文件可以直接用钢筋检测里面的替换

或者在群文件下载RFSong多数据集联合训练文件夹（因为多数据集联合训练就统一用的VOC2007格式）

运行train_RFB

与RFB不同的是

代码更方便进行自己设计网络
网络非常轻量级只有3.1 MB
网络速度相比RFB有显著提升,能够达到200FPS

更新

在VOC0712+COCO+PRW+SYSU总共大概十万张图片的数据集上进行了训练，模型代码和权重都放在了群文件，实测效果得到了显著提升，有需要的可以直接加群下载使用。

欢迎加群交流，云深不知处-目标检测 763679865

rfsong-7993's People

Contributors

Stargazers

Watchers

rfsong-7993's Issues

200fps?

与给出的200fps相差很多，本人在新TianX上只有100fps，请问你是在哪种型号的显卡跑的？

对网络进行了从头训练问题？

对网络进行了从头训练遇到过：
1.网络难于收敛，
2.如果一直不收敛，怎么去调整参数

那个群号搜索不到，能发下VOC0712+COCO+PRW+SYSU训练的权重吗

RFSong7993-to-caffe

how to convert RFSong7993(.pth) to caffemodel?

建议删掉/utils/nms/ 路径下的cpu_nms.c gpu_nms.cpp这2个文件，可能会导致nms编译出错。

i

train hyper parameter

1.Which lr is suitable for 4 gpus?
2.How many epochs are needed to get a good model for the dataset you mentioned?
I use you default setting to train the model for 220 epoches,the result is really weird.The boxes are everywhere.

.make.sh, python testRFB, error

/utils/nms_wrapper.py", in
from .nms.cpu_nms import cpu_nms, cpu_soft_nms
ImportError: No module named nms.cpu_nms

ValueError: not enough values to unpack (expected 2, got 1)

python train_RFB.py
Loading Dataset...
Traceback (most recent call last):
File "train_RFB.py", line 166, in
train()
File "train_RFB.py", line 88, in train
dataset = VOCDetection(VOCroot, train_sets, preproc(img_dim, rgb_means, p), AnnotationTransform())
File "/home/rencong/RFBNET/SSD-YOLO/data/voc0712.py", line 120, in init
img_id, value = line.split()
ValueError: not enough values to unpack (expected 2, got 1)

。

Recommend Projects

React

A declarative, efficient, and flexible JavaScript library for building user interfaces.
Vue.js

🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
Typescript

TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
TensorFlow

An Open Source Machine Learning Framework for Everyone
Django

The Web framework for perfectionists with deadlines.
Laravel

A PHP framework for web artisans
D3

Bring data to life with SVG, Canvas and HTML. 📊📈🎉

Recommend Topics

javascript

JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
web

Some thing interesting about web. New door for the world.
server

A server is a program made to process requests and deliver data to clients.
Machine learning

Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
Visualization

Some thing interesting about visualization, use data art
Game

Some thing interesting about game, make everyone happy.

Recommend Org

Facebook

We are working to build community through open source technology. NB: members must have two-factor auth.
Microsoft

Open source projects and samples from Microsoft.
Google

Google ❤️ Open Source for everyone.
Alibaba

Alibaba Open Source for everyone
D3

Data-Driven Documents codes.
Tencent

China tencent open source team.