当前仓库属于暂停状态,部分功能使用受限,详情请查阅 仓库状态说明
1 Star 0 Fork 89

不死鸟玉山禾 / Argus
暂停

forked from 张红元 / Argus
暂停
 
加入 Gitee
与超过 1200万 开发者一起发现、参与优秀开源项目,私有仓库也完全免费 :)
免费加入
克隆/下载
贡献代码
同步代码
取消
提示: 由于 Git 不支持空文件夾,创建文件夹后会生成空的 .keep 文件
Loading...
README
LGPL-3.0

项目地址:https://gitee.com/windandwine/Argus 转载请注明出处

一、项目简介

fine-tune YOLO v3 + FaceNet进行人脸识别,辨别。

输入图片说明

1. 项目结构

--data

--|------baseface 图片、根据这些图片训练的128d向量,以及文件夹与人名的映射文件

---------|------0 第一个人的图片tag=0

---------|------1 第二个人的图片tag=1

......

---------|------n 第n个人的图片tag=n

---------|------map.txt 文件夹与人名的映射,依次放即可

---------|------vector.csv 根据这些图片得到的128维向量以及其类别(文件夹名)

--|------weights_facenet 权重(facenet)

--|------weights_yolo 权重(fune-tuning的yolo v3)

--|------weights_svm 模型(根据csv文件训练的svm模型)

--|------face-names 预测类别,默认即可

--|------yolo_anchors.txt 训练yolo v3时聚类得到的anchors框

--net yolo和facenet的网络

--preprocessing 预处理工具,下文的使用方法前期步骤都需要在这里运行

--setting 模型参数,可根据需要修改

--utils 封装的一些方法

--test.py 主体方法

2.权重文件

yolo v3是基于论文作者模型,在Wider Face数据上fine-tuning的,可以到这里下载,密码qli4

facenet权重

  1. model-20170512-110547,请到这里下载,密码9x2l
  2. model-20180408-102900,请到这里下载,密码jjf1
  3. model-20180402-114759(推荐),请到这里下载,密码u91l

3.yolo v3

YOLO v3的详细预测和训练,可到本人另一个项目YOLO_v3_tensorflow了解。

4.踩的坑

  1. fine-tune yolo v3时使用的是wider face数据集,其中有两个标注框是宽度或高度为0的,错误标注,筛选的时候需要去除掉这两个标注框,否则nms会报除0的异常。
  2. 训练svm时,需要标准化,预测时需要用同参数标准化再预测,否则svm预测结果都相同。
  3. 之前使用model-20170512-110547模型,输出128d向量,效果不好,换成model-20180402-114759模型,输出512d向量,效果有所提升。
  4. 每个人15张脸部图片,训练svm效果一般,可以增大样本量或者使用一些svm的tricks。

三、使用方法

项目需要安装以下包

numpy
pandas
opencv-python
scikit-learn
tensorflow
pillow

1.制作自己的人脸数据集

截取需要识别的人物的脸部图片,一人一个文件夹,文件夹名称从0开始依次累加,放在路径data/base_face下。

输入图片说明

并修改data/map.txt,以空格分隔,下标和人脸文件夹名一一对应(map.txt中第一条下标为0,对应data/base_face/0文件夹)。

2.使用工具将图片转换成向量并存储

运行preprecessing/pre_tools.py内的save_vector_csv(),将图片使用facenet转换为128d或512d向量,并存储为data/base_face/vector.csv中。

3.训练svm分类器

基于已经储存的vector.csv文件,进行标准化后,运行preprecessing/pre_tools.py内的train_face_svm(),使用scikit-learn训练svm模型,并储存在data/weights_svm/svm.pkl中。

4.开始测试

4.1图片测试

放在data/test_img下,将setting/yolo_args.py中的detect_object改为imginput_image 改为图片路径。

运行根目录下的test.py文件。

另外如果要存储检测后的图片,将setting/yolo_args.py中的output_image改为要存储的路径,并保证各层级文件夹存在

4.2视频测试

放在data/test_video下,将setting/yolo_args.py中的detect_object改为videoinput_video 改为视频路径。

运行根目录下的test.py文件。

另外如果要存储检测后的视频,将setting/yolo_args.py中的output_video改为要存储的路径,并保证各层级文件夹存在

GNU LESSER GENERAL PUBLIC LICENSE Version 3, 29 June 2007 Copyright (C) 2007 Free Software Foundation, Inc. <http://fsf.org/> Everyone is permitted to copy and distribute verbatim copies of this license document, but changing it is not allowed. This version of the GNU Lesser General Public License incorporates the terms and conditions of version 3 of the GNU General Public License, supplemented by the additional permissions listed below. 0. Additional Definitions. As used herein, "this License" refers to version 3 of the GNU Lesser General Public License, and the "GNU GPL" refers to version 3 of the GNU General Public License. "The Library" refers to a covered work governed by this License, other than an Application or a Combined Work as defined below. An "Application" is any work that makes use of an interface provided by the Library, but which is not otherwise based on the Library. Defining a subclass of a class defined by the Library is deemed a mode of using an interface provided by the Library. A "Combined Work" is a work produced by combining or linking an Application with the Library. The particular version of the Library with which the Combined Work was made is also called the "Linked Version". The "Minimal Corresponding Source" for a Combined Work means the Corresponding Source for the Combined Work, excluding any source code for portions of the Combined Work that, considered in isolation, are based on the Application, and not on the Linked Version. The "Corresponding Application Code" for a Combined Work means the object code and/or source code for the Application, including any data and utility programs needed for reproducing the Combined Work from the Application, but excluding the System Libraries of the Combined Work. 1. Exception to Section 3 of the GNU GPL. You may convey a covered work under sections 3 and 4 of this License without being bound by section 3 of the GNU GPL. 2. Conveying Modified Versions. If you modify a copy of the Library, and, in your modifications, a facility refers to a function or data to be supplied by an Application that uses the facility (other than as an argument passed when the facility is invoked), then you may convey a copy of the modified version: a) under this License, provided that you make a good faith effort to ensure that, in the event an Application does not supply the function or data, the facility still operates, and performs whatever part of its purpose remains meaningful, or b) under the GNU GPL, with none of the additional permissions of this License applicable to that copy. 3. Object Code Incorporating Material from Library Header Files. The object code form of an Application may incorporate material from a header file that is part of the Library. You may convey such object code under terms of your choice, provided that, if the incorporated material is not limited to numerical parameters, data structure layouts and accessors, or small macros, inline functions and templates (ten or fewer lines in length), you do both of the following: a) Give prominent notice with each copy of the object code that the Library is used in it and that the Library and its use are covered by this License. b) Accompany the object code with a copy of the GNU GPL and this license document. 4. Combined Works. You may convey a Combined Work under terms of your choice that, taken together, effectively do not restrict modification of the portions of the Library contained in the Combined Work and reverse engineering for debugging such modifications, if you also do each of the following: a) Give prominent notice with each copy of the Combined Work that the Library is used in it and that the Library and its use are covered by this License. b) Accompany the Combined Work with a copy of the GNU GPL and this license document. c) For a Combined Work that displays copyright notices during execution, include the copyright notice for the Library among these notices, as well as a reference directing the user to the copies of the GNU GPL and this license document. d) Do one of the following: 0) Convey the Minimal Corresponding Source under the terms of this License, and the Corresponding Application Code in a form suitable for, and under terms that permit, the user to recombine or relink the Application with a modified version of the Linked Version to produce a modified Combined Work, in the manner specified by section 6 of the GNU GPL for conveying Corresponding Source. 1) Use a suitable shared library mechanism for linking with the Library. A suitable mechanism is one that (a) uses at run time a copy of the Library already present on the user's computer system, and (b) will operate properly with a modified version of the Library that is interface-compatible with the Linked Version. e) Provide Installation Information, but only if you would otherwise be required to provide such information under section 6 of the GNU GPL, and only to the extent that such information is necessary to install and execute a modified version of the Combined Work produced by recombining or relinking the Application with a modified version of the Linked Version. (If you use option 4d0, the Installation Information must accompany the Minimal Corresponding Source and Corresponding Application Code. If you use option 4d1, you must provide the Installation Information in the manner specified by section 6 of the GNU GPL for conveying Corresponding Source.) 5. Combined Libraries. You may place library facilities that are a work based on the Library side by side in a single library together with other library facilities that are not Applications and are not covered by this License, and convey such a combined library under terms of your choice, if you do both of the following: a) Accompany the combined library with a copy of the same work based on the Library, uncombined with any other library facilities, conveyed under the terms of this License. b) Give prominent notice with the combined library that part of it is a work based on the Library, and explaining where to find the accompanying uncombined form of the same work. 6. Revised Versions of the GNU Lesser General Public License. The Free Software Foundation may publish revised and/or new versions of the GNU Lesser General Public License from time to time. Such new versions will be similar in spirit to the present version, but may differ in detail to address new problems or concerns. Each version is given a distinguishing version number. If the Library as you received it specifies that a certain numbered version of the GNU Lesser General Public License "or any later version" applies to it, you have the option of following the terms and conditions either of that published version or of any later version published by the Free Software Foundation. If the Library as you received it does not specify a version number of the GNU Lesser General Public License, you may choose any version of the GNU Lesser General Public License ever published by the Free Software Foundation. If the Library as you received it specifies that a proxy can decide whether future versions of the GNU Lesser General Public License shall apply, that proxy's public statement of acceptance of any version is permanent authorization for you to choose that version for the Library.

简介

基于YOLO v3、FaceNet和SVM的人脸检测识别系统 展开 收起
Python
LGPL-3.0
取消

发行版

暂无发行版

贡献者

全部

近期动态

加载更多
不能加载更多了
Python
1
https://gitee.com/yushanhe/Argus.git
git@gitee.com:yushanhe/Argus.git
yushanhe
Argus
Argus
master

搜索帮助