当前位置:网站首页>Learning summary (about deep learning, vision and learning experience)
Learning summary (about deep learning, vision and learning experience)
2020-11-08 09:45:00 【osc_sqtwbty4】
It's today 2020 year 11 month 6 Japan , It's been a month since I came to Shanghai , I want to write a summary of my study , Then I started to try new ways of learning .
Catalog
of OpenCV+python Computer vision image processing learning
About deep learning basic learning
2020 year 3 Computer vision begins in January , Contact keras frame , A lot of them didn't understand at that time , Bit by bit , Finally, I successfully completed my first visual project, which is my graduation project 《 Design and implementation of facial expression recognition system based on convolutional neural network 》. Now simply recall the learning process at that time , By the way, I'd like to summarize a little bit of experience , I hope it will be helpful to the students at the beginning , Come on together ( Some learning videos may not be very classic , But it was my own choice, consciously or unconsciously , Therefore, it is only for reference ).
Learning material sharing
of python Study
First contact python Is in 2018 year 10 month , At that time, a person was painting videos in the undergraduate laboratory during the National Day holiday , Make notes , Crazy input . What we learn is 《[ Little turtle ] Basic learning Python》, Unfortunately, it was not fully used at that time CSDN Blog , I wrote down all the notes in the Word file , A lot of records , It's a bit messy . It's a bit stupid to take notes at first , A lot of them , At the same time, it also taps the code , Keep trying 、 Errors and Solutions . Sometimes encounter fun will also write their own code to amuse themselves ( This is a screenshot of the circle of friends ).
B Station links :https://b23.tv/kLRXOX ( Partial contents and screenshots )
of OpenCV+python Computer vision image processing learning
Originally in B I've learned from you once , A video to get me started 《OpenCV+python Computer vision image processing 》, I started taking notes by hand, but I found it was too slow , We found that the video code was pasted in the video introduction , Download and learn quickly . Follow the video to configure the required environment , Follow the video and comment on the code provided , Write your own understanding . It used to be using Typora This kind of Markdown Editor to take notes , After a period of time, it should be that I didn't master it well ( The picture reprint aspect appears the problem, has been dragging has not solved ), It's right there CSDN To take notes ( See the screenshot of the column below ).
B Station links :https://www.bilibili.com/video/BV1QJ411W7SS?p=1 ( Partial contents and screenshots )
Screenshot of column ( Some notes at that time are arranged here , Later, I learned something about it and kept it , It may be a bit messy in some places , Your smile .)
There are also some quality book resources as follows :
link :https://pan.baidu.com/s/1rAzNRaJSe8EwxWVVTCQp6g Extraction code :i924
On neural network learning
We need to be a learning power every day , At that time, I saw what was said by Chen Bin, a teacher of Peking University 《 Artificial intelligence and information society 》( Partial theory , Consider getting started ),
I just started studying in a powerful country , Later it was found that there was no double speed , A little sad , Went directly to the corresponding MOOCS to learn , It's great to find MOOCS , You can also download the corresponding PPT, I wrote some notes in the blog column 《 Notes on learning neurons and neural networks 》
MOOC link :https://www.icourse163.org/course/0809PKU037-1003471009
About deep learning basic learning
Partial theory
I learned from Wu Enda , It was B Station learning , It was later found that Netease cloud classroom has courses . Overall, I felt very good , The teacher said it over and over again , Suitable for entry , Partial theory . Just started taking notes by taking screen shots , It's too slow to find out , I accidentally found that Mr. Huang haiguang wrote a complete set for this course 《DeepLearning.ai In depth learning course notes 》 And always updated , What's up to date with me now 5.7 edition . I also made my own notes based on the video presentation and course notes , Although many of them are copy and paste , But I still want to do it , Do some highlights and occasional summaries , This is more profound .( Address of Wu Enda's in depth learning notes column :https://blog.csdn.net/dujuancao11/category_9871211.html , It is suggested that you watch the video and take notes , I write a simple look at the line , I'm for the convenience of my own review .)
Netease cloud classroom link :https://mooc.study.163.com/smartSpec/detail/1001319001.htm
Cover of supporting notes and screenshots of some contents :
link :https://pan.baidu.com/s/142Lj7n5hneg-VgjgkM7fvg Extraction code :tj40
Biased practice
After learning Wu Enda's deep learning course , I learned 《 Hands-on deep learning Pytorch edition 》, This time, I have a deeper understanding of some basic things in-depth learning , And it's more about using Pytorch Framework to implement , Not only deepen the understanding of the foundation of deep learning , And it's easy to learn Pytorch Basic syntax .
Some catalogues and Book screenshots
Corresponding source code :https://github.com/ShusenTang/Dive-into-DL-PyTorch
Network disk link :https://pan.baidu.com/s/1HdATN4RZxdEDZbCtSrPUdw Extraction code :z66y
On frame learning
Keras
The first frame I used was keras( You can put Keras regard as tensorflow After the encapsulation API), There is a corresponding official document, and there is one PDF file .
Keras Chinese document :https://keras.io/zh/
Screen capture of document section :
Network disk link :https://pan.baidu.com/s/1BVBjLTi0UGtP_VVkIzwVLw Extraction code :1145
tensorflow
Simple contact , Use Keras When you need it , I've written a few blog posts about it before .
tensorflow Entry variable constant :https://blog.csdn.net/dujuancao11/article/details/104896330
3tensorflow arithmetic :https://blog.csdn.net/dujuancao11/article/details/105051971
The biggest headache should be tensorflow-gpu Configuration of Well , In this post :《 Virtual environment installation 》https://blog.csdn.net/dujuancao11/article/details/107468687 ( There are a lot of things , It may be a bit messy )
You can also read this one :《 of Python Simple installation of some commonly used software ( Baidu disk installation package link attached )》https://blog.csdn.net/dujuancao11/article/details/109394751
And in Problems in configuration :《tf-gpu( testing 、np.dtype([(“qint8“, np.int8, 1)]) 、 not compiled to use: AVX2、CUDNN_STATUS_ALLOC_FAILED)》https://blog.csdn.net/dujuancao11/article/details/109111198
Pytorch
In fact, this framework is mainly based on 《 Hands-on deep learning Pytorch edition 》 In contact with , There is no specialized study yet , But there are also learning documents 《PyTorch Official course in Chinese 》.
Network disk link :https://pan.baidu.com/s/1k9XrhuPI3nFJP8tweNrLsQ Extraction code :w6ve
Summary
The above recommendations PyTorch
The above are some of the materials I used in the learning process , Complete Baidu disk link :https://pan.baidu.com/s/1ygkz4rv2nZQrnn86i4Weyw Extraction code :5u7y
Thinking method
Based on learning
- Based on learning : promote Search for information The ability of , Learn to learn to stand on the shoulders of giants , Learn to search the corresponding source code 、 E-books and PPT etc. .
Some sources :CSDN、B standing 、 WeChat official account 、 You know 、 Simple books ( Serious exploration will find learning power is also good , For example, new developments in a certain field 、 The new policy , Facilitate better decision making .)
I have to admit that there are many excellent people around here doing official account. B Stop video , For example, I love WeChat official account. : Handsome play with programming 、 Gongzilong 、OAOA( My high school classmates ) etc. ,CSDN Blogger :
Bubbliiiing( It's really Example ).
The paper : I read too little , The next goal is to read the paper well . - note : It's necessary to take notes yourself , Choose your own way to take notes , For example, I'm used to it CSDN Blog notes , Although sometimes it's just a simple copy and paste , But I'm going to do it again , Bold and red in some places , Write your own thoughts in some places 、 Summary or doubt . Take notes in time , Wrote a lot of blogs , Start to be strict with yourself , Although it is still far from the target .
Possible problems :
1) To solve a problem , I read too much information , But there are some gains , No time to take notes , You can choose the key points to take notes , You can also focus on a time to write .
2) Blog errors may occur , Because they are basically self-taught , It is inevitable that there will be a lack of understanding , Boldly remember , Continuous updating and modification .( If there is a mistake in my blog , Welcome to point out , Come on together . But don't be too demanding , Time and energy are limited , I write as well as I can . Thank you for your support !)
3) There are some relatively new ideas , With protection in mind , May be set to private . - communication : And mentors 、 There are also excellent students to communicate with each other , A lot of inspiration , Cherish everything now .
solve the problem
- State of mind : I think we must have a good attitude , Download a software 、 Configuration environment 、 Write code 、 Debugging and running problems are too normal , Don't be afraid of trouble , Facing problems , Learn to search and solve by yourself . A lot of times, there's a problem , A solution others can , You can't and it's normal , Just try more ( It's right to have a good time ).
- Get from the problem inspire :
example 1: I didn't know anything at the beginning of deep learning , Direct use CPU Training Fer2013 Data sets and Inception The Internet , Training speed moving , Too slow , Only later did I know it could be used GPU Speed up ( Have you read this article :https://zhuanlan.zhihu.com/p/30751437), Also know the need to support training graphics card can ( Graphic card calculator https://developer.nvidia.com/cuda-gpus), Later he learned to configure Tensorflow-gpu.
example 2: A little accident , There is no sound in the video recording , Thinking about putting the PPT Intercept it ,1 An hour of video , It's too slow to intercept one by one , A smart machine opencv Try it . See my blog for details :https://blog.csdn.net/dujuancao11/article/details/109404108
New ideas and inspiration
- Take part in more Academic Conference , Last time CGCKD( Organized by Shanxi University 2020 Chinese Symposium on Granular Computing and knowledge discovery ( The 20th China Rough Conference on set and soft computing 、 The 14th China Granular Computing Conference 、 The eighth academic conference on three branches of decision making )), The biggest learning is : At present, artificial intelligence is based on computation , Next stage based on memory . I also feel the teachers' love and passion for academic , An example of learning , I hope that one day we can become as excellent scientific researchers as they are , Do things in a down-to-earth manner .
- actual : We should be good at connecting what we have learned in real life , It has a certain sensitivity , I think the ultimate goal of scientific research is to serve the society , Service provider itself .
- Be good at understanding some policies or disciplines dynamic , Convenient and better decision making .
expectation
Writing this blog is also to end this stage of learning , Try to start a new learning method .
- Work and rest more regularly , Keep early hours .
- Be more self disciplined , Exercise on running ( Exercise is also creating conditions for scientific research ).
- More planning for study .
- Be more modest ( I've always been cai, It's far from the target ).
- hope xiyu Deo gratias , Come on together !
Summary
That's the end of sharing , A little experience , In fact, it is “ come clean ” The share of , I'm under a lot of pressure , Because I'm exploring a lot of places , Not very well . But I'm still willing to sort out the fragmented learning links before , Spend hours writing this blog post . Thank you for everything now , Appreciate support and like ( As of now 484 Fans ), I'll work harder to blog .
I spent a whole month in graduate school , Like school , Teachers met 、 Senior brothers and sisters and classmates are very good , Happy every day , Some are studying hard , Slowly groping , Make friends seriously . Although most of the time is busy , But because there are a group of lovely people around and become happy and full .
Although I don't know what will happen in the future , But will always be enthusiastic , Strive to move forward in the exploration .
To know and then to know , Be sure to be quiet , Be quiet and then be safe , You can worry about it , Worry before you get . Everything has its roots and ends , There is an end to it . know what should precede and what should follow , It's a short cut .
Please give me more advice , Come on !
版权声明
本文为[osc_sqtwbty4]所创,转载请带上原文链接,感谢
边栏推荐
- Architect (November 2020)
- “智能5G”引领世界,数位智能网优+5G能带来什么?
- Astra: Apache Cassandra的未来是云原生
- 函数周期表丨筛选丨值丨SELECTEDVALUE - 知乎
- 学习小结(关于深度学习、视觉和学习体会)
- Flink's sink: a preliminary study
- 进程、线程和协程的区别
- PCIe enumeration process
- 2020-11-05
- Close to the double 11, he made up for two months and successfully took the offer from a large factory and transferred to Alibaba
猜你喜欢
AMD Zen3首发评测:频率超5GHz,IPC提升不止19%,这次真的Yes了 - 知乎
Basic concepts of computer network (5) basic principles of local area network
How can a technician take over a complex system?
Rust: command line parameter and environment variable operation
Python loop distinction (while loop and for loop)
Review the cloud computing application scenarios you didn't expect (Part 1)
Adobe Prelude / PL 2020 software installation package (with installation tutorial)
来自不同行业领域的50多个对象检测数据集
攻防世界之web新手题
分布式共识机制
随机推荐
Can you do it with only six characters?
AMD Zen3首发评测:频率超5GHz,IPC提升不止19%,这次真的Yes了 - 知乎
临近双11,恶补了两个月成功拿下大厂offer,跳槽到阿里巴巴
Solve the problem of rabbitmq message loss and repeated consumption
Face recognition: attack types and anti spoofing techniques
Visual Studio 2015 未响应/已停止工作的问题解决
Shiyou's numerical analysis assignment
5g/4g工业无线路由器
iOS上传App Store报错:this action cannot be completed -22421 解决方案
Julia 是如何风靡起来的?
2020-11-05
What details does C + + improve on the basis of C
抖音直播监控Api:随机推荐
It's 20% faster than python. Are you excited?
TCP协议如何确保可靠传输
Application of bidirectional LSTM in outlier detection of time series
软件测试培训班出来好找工作么
Seven features of Python 3.9
nvm
413【毕设课设】基于51单片机无线zigbee无线智能家居光照温湿度传输监测系统