当前位置:网站首页>Some problems in face recognition testing with facenet source code
Some problems in face recognition testing with facenet source code
2022-07-19 05:45:00 【Just do it! ට⋆*】
1. First download facenet Source code :https://github.com/davidsandberg/facenet
2. function facenet
use pycharm open , Download it tensorflow1.7 edition , Corresponding python Version is 3.6, There's a hole here , install tensoorflow1.7.0 When it comes to numpy Inconsistent with other library versions , It needs to be reinstalled step by step according to the prompts .
Installation Library :conda install Library name == edition
Delete Library :conda uninstall Library name
3. download flw Data sets
link : http://vis-www.cs.umass.edu/lfw/
After downloading , Unzip to data\lfw_data\lfw, If there is no folder, create it yourself
4. Preprocess photos
The face photos downloaded here are 250250 Pixels , Need to change to 160160 Pixels ,
stay pycharm Open in align_dataset_mtcnn.py file , To configure parameters:
Download datasets Store the processed data set Processing settings ( Create a file based on the path )
D:\yanyi\project_process\facenet-master\src\data\lfw_data\lfw D:\yanyi\project_process\facenet-master\src\data\lfw_data\lfw_160 --image_size 160 --margin 32 --random_order --gpu_memory_fraction 0.25
After processing the photos , Can be in lfw_160 See the processed file in the file 
stay data A paris.txt file
5. Test the model
Mode one : download facenet Training models
Click the download link
What is used here is 20180408-102900 Model under file , Copy this file to facenet\src\models Under the table of contents .
Here I first use pycharm Add parameter :
The path where the dataset is located The path of the model
Such as D:\yanyi\project_process\facenet-master\src\data\lfw_data\lfw_160 D:\yanyi\project_process\facenet-master\src\models\20180408-102900
Add parameters to parameters, function validate_on_lfw.py After the discovery
There has been a lack data\paris.txt file error , No solution has yet been found , But use cmd After operation , Running successfully !
cd To src Under the document :
python validate_on_lfw.py D:\yanyi\project_process\facenet-master\src\data\lfw_data\lfw_160 D:\yanyi\project_process\facenet-master\src\models\20180408-102900

Recognition accuracy reaches 0.97, Test success
There are some mistakes , Not yet
边栏推荐
- Could not locate zlibwapi.dll. Please make sure it is in your library path
- 基于opencv的检测人脸,并截取人脸
- Wxml template syntax in wechat applet
- Unable to determine Electron version. Please specify an Electron version
- List and map
- 记录:YOLOv5模型剪枝轻量化
- 使用OpenCV、ONNXRuntime部署YOLOV7目标检测——记录贴
- 正则替换group(n)内容
- Simple application of COAP in andorid
- 微信小程序之计算器
猜你喜欢

Unable to determine Electron version. Please specify an Electron version

Pointnet++代码详解(一):farthest_point_sample函数

Pointnet++代码详解(二):square_distance函数

记录:YOLOv5模型剪枝轻量化

3. Neusoft cross border e-commerce data warehouse project architecture design

基于四叉树的图像压缩问题

PyTorch学习笔记【3】:使用神经网络拟合数据

Calculator of wechat applet

Ambari 2.7.5 integrated installation hue 4.6

10. DWD layer construction of data warehouse construction
随机推荐
软件过程与管理复习(七)
安卓实现真正安全的退出app
Composants communs des applets Wechat
Pointnet++ code explanation (I): farthest_ point_ Sample function
PyTorch学习笔记【4】:从图像学习
微信小程序密码显示隐藏(小眼睛)
图片的大小限制显示
MySQL learning notes (4) - (basic crud) operate the data of tables in the database
Common components of wechat applet
1. Dongsoft Cross - Border E - commerce Data Warehouse Requirement specification document
使用OpenCV、ONNXRuntime部署YOLOV7目标检测——记录贴
INRIAPerson数据集转化为yolo训练格式并可视化
Pointnet++代码详解(二):square_distance函数
SGM: Sequence Generation Model for Multi-Label Classification(用于多标签分类的序列生成模型)
Write a timed self-test
DEEP JOINT TRANSMISSION-RECOGNITION FOR POWER-CONSTRAINED IOT DEVICES
cuda11.0的pytorch安装小计
写一个定时自检
List and map
Wxml template syntax in wechat applet