Image Identification.
First, set up the environment
to achieve python-based TensorFlow
installation 1.anaconda of
reference: HTTPS: //jingyan.baidu.com/article/f0062228503d2afbd3f0c8fe.html
2. Visualization Tool: jupyter notebook literate programming is conducive to data analysis
Reference: https: // www.jianshu.com/p/97fa4ed3edbc
open: after installing anacoda enter cmd in
jupyter notebook
clicks in new browser pop-up interface to start writing code.
Close ways: you can click on shutdown in the notebook. Shut down the entire service: In the original terminal press Control + c two
code units: Here is where you write the code by pressing Shift + Enter to run the code, the results displayed below this unit.
The left-hand unit has In [1]: a sequence tag code is easy for people to see the execution order.
Markdown unit: here on the text editor, syntax specification uses markdown, you can format text, insert links, images and even mathematical formulas.
Shift + Enter run using the same markdown unit to display formatted text.
Linux-like Vim editor, the notebook also has two modes:
Edit mode: Edit text and code. Select unit and press Enter to enter the edit mode, the left unit case green vertical line.
Command Mode: keyboard shortcut command for executing input. By Esc to enter command mode, the left unit case blue vertical line.
If you want to use shortcut keys, first press Esc to enter command mode, then press the corresponding key to achieve the operation of the document.
Such as switching unit into a code (Y) or markdown unit (M), an increase or a unit (B) beneath the module. View all shortcut command can press H.
Second, the use of image recognition keras
official document: HTTPS: //keras.io/zh/
https://keras-cn.readthedocs.io/en/latest/other/optimizers/
1.python3.7 back to 3.6
Open anaconda prompt -> input the install Python = Conda 3.6 -> Y
2. Install tensorFlow2.0:
① installed after downloading the corresponding version of the anaconda, the command installation tensorflow2.0 (such as the need to replace the line after the official version):
pip install tensorflow==2.0.0-beta1
②python virtual environment
Create a new virtual environment:
conda create -n tensorflow python=3.6
View existing virtual environment:
conda env list
Switched virtual environment (name when tensorflow create for yourself, here tensorflow):
activate tensorflow
In this open jupyter notebook
In this Open Notebook jupyter
③ create a new file python3
import tensorflow as tf print(tf.__path__) Print (TF .__ version__) // Note that two bars
④ Detection tensorflow whether the installation is successful
pip show tensorflow #ttributeError: module 'tensorflow' has no attribute '_version_'
No solution?
2.0.0-beta1 #
⑤ install opencv environment and keras
pip install hard
pip install opencv-python
Based on structures and visualization of CNN keras classification model (with detailed code)
https://blog.csdn.net/u011268787/article/details/79891284
Three, keras learning framework to develop
artificial small example (a) https://blog.csdn.net/xvshu/article/details/79026234
Recognition Based keras + openCV (b) https://blog.csdn.net / xvshu / article / details / 78863430
face detection when people pay attention to modify the good data / haarcascade_frontalface_alt2.xml path to
learning materials: python depth study [US] Fulangsuowa · Chaulet the translation of Zhang Liang
1.kaggle game (data analysis, mathematical modeling) https://blog.csdn.net/lezaimemecha/article/details/78906337
2. The basic linear algebra routines (BLAS, basic linear algebra subprograms)
four,
1. Using a small tool to activate Windows
https://pan.baidu.com/s/1JJ39b6kk34J3wyATvPr7ZA
2. Install Chinese Xingkai font (you can go directly to the Free Software search you can also find in my Baidu network disk inside) https://pan.baidu.com/s/1WW930cM6CeODdPRKEhVIDA
copy .ttf file to
Fifth, pattern recognition
1. Machine Learning generic templates:
① problem definition, data collection set
② selection of indicators to measure the success of
③ to determine the assessment method
④ ready data
⑤ develop a better model than the benchmark
⑥ expand the scale model: model fitting developed over
⑦ model regularization and adjusting hyperparameter
2. Ali cloud free interfaces
① multimedia AI - Artificial Intelligence Product Ali cloud - Multimedia AI (beta) release
https://www.aliyun.com/product/news/detail?&msctype=email&mscareaid=cn&mscsiteid=cn&mscmsgid=2260119110502326459&id=14341&source_type= out_edm_feature_20191105
② pictures bad scene recognition
https://ai.aliyun.com/lvwang/imglive?spm=5176.12418109.h2v3icoap.229.4ecd29b3e40hpL
③openpose framework for achieving human skeleton detection
https://github.com/tensorlayer/openpose-plus