当前位置:网站首页>The best Chinese open source class of vision transformer, ten hours of on-site coding to play with the popular model of Vit!
The best Chinese open source class of vision transformer, ten hours of on-site coding to play with the popular model of Vit!
2022-06-26 05:05:00 【Paddlepaddle】
《Vision Transformer Punch camp 》 It's officially Online ! Senior researcher of the deep learning Research Office of Baidu Research Institute Dr. Zhu Wu The whole course , Six professional teaching assistants answer questions throughout the course , Ten lessons to show you around ViT Explosion model ! Teaching in Chinese , Fully open source !
Class time :11 The moon 23 Japan — 12 month 2 Japan , Every night 20:30 Start live teaching on time ~
Course link ( The playback 、 Supporting practice ):https://aistudio.baidu.com/aistudio/course/introduce/25102
official Q Group ( Answering question 、 Technical communication ): Search for group numbers 920200490 Or scan the QR code below to reply to the keyword to get the way to join the group ~

Transformer since 2017 After it was proposed in , From sweep NLP The scenery of the field is the same , To get caught up in a field of CV In the voice of questioning the effectiveness of the mission , Not long ago, it showed a direct approach in a number of image tasks CNN Excellent performance of as well as ICCV2021 best paper:Swin Transformer It's hot ,Transformer Step by step CV The field is brilliant !
Therefore ,Transformer、Vision Transformer、QKV、Attention is all you need And so on Ignite the academic circle !

As a developer , While applauding this academic Superstar , why not Get on the bus in time , understand Encoder decoder QKV What is it 、 All over the Internet Swin Transformer What did you do right ?Transformer Why can you sweep NLP, again How the CV The role played in the task ? Current CV Great God's new work MAE Algorithm And how it works ?
flowers 10 For an hour Really understand the technical value behind these hot words , Make these hot money by yourself ViT Algorithm , Even apply it to my current work and scientific research , Become the next top hit !

《ViT Ten lectures 》Is All You Need!
11 month 23 Japan ( Tuesday ),AI Top reviewer 、 Dr. Zhu Wu, senior researcher of Baidu deep learning institute Live lecture 《Vision Transformer Ten lectures 》. Every night 1 Hours 、 continuity 10 God , Dr. Zhu Wu will take you Play from scratch ViT Algorithm !
Vision Transformer Comprehensive explanation of cutting-edge technologies : from ViT Classical algorithm to the latest academic frontier , From technical principles to implementation details , ViT, Swin, DETR Take you to master the new paradigm of visual algorithm one by one .

Paper analysis + Line by line Coding、 On site blackboard writing , Hand formula : Analyze the key points of the thesis in detail , Half of each class will take you to write code on the spot ~ Take everyone to realize their own ViT Model , Yes The small white It's also very friendly .

The entire carry Get along well with ViT: Course content Fully open source 、 The entire Live Chinese teaching 、 High level teaching assistants answer questions in the community 、 Practical tasks with different difficulties are provided to meet the learning needs of each stage , And provide free GPU Computational support !
Join the open source technology team , Become an academic suite with the leaders of the Research Institute PPViT At the heart of contributor: common Reduce ViT Model implementation threshold , Improve the productivity of scientific research and learning .(https://github.com/BR-IDL/PaddleViT )
You will reap
Whether you are new to deep learning , Still doing scientific research , No matter you are CV Want to turn NLP, still NLP Want to CV, Or you want to play with the latest visual technology 、 Send a paper , Through the study 《Vision Transformer Ten lectures 》, You will gain the ability to turn the model diagram in the paper into lines of code , Say goodbye to simple git clone And transfer !
About Instructor
Dr. Zhu Wu , Senior researcher of deep learning laboratory of Baidu Research Institute 、 Flying propeller PPDE( Developers, technologists )、AI Top reviewer . Have first-line working experience in Silicon Valley in the United States , Now in charge of deep learning and visual algorithm research and development in Baidu Research Institute . He has published many high-level papers , Many times in recent years CVPR, ICCV, ECCV Wait for the international competition at the summit to win top Ranking , Person to person nickname “ The God who knocks code by hand ”.
Syllabus
Lesson one
theory :Hello, Vision Transformer!
practice :Warmup: Model building and training
The second is
theory : from Transformer To Vision Transformer
practice : Get along well with Tensor operation , Begin to build ViT
The third is
theory : Look at your , I see my detailed attention
practice :Multi-Head Self Attention
Lesson four
theory : Build your first ViT Model
practice : How to achieve ViT Model
The first five lectures
theory :ViT The model is built , How to train efficiently ?
practice : Actual combat model building and training
About 6
theory : What is? Window Attention?
practice : The attention mechanism on the image window
Lecture 7
theory : big-name Swin Transformer
practice : Realize your second ViT Model
Lesson eight
theory :Conv and Transformer The combination of
practice : From the framework source code to see how to achieve data loading
About 9
theory : Take you to the frontier algorithm : Visually BERT
practice : Skills of model training
Lecture 10
theory : New paradigm of detection algorithm -DETR
practice : actual combat ViT The whole process of training and testing
Marriott gift , Just waiting for you
Participate in the course 、 Pay tribute to open source , You can get :PaddlePaddle Official certificate of completion 、Marshall ACTON II BLUETOOTH Wireless subwoofer 、HHKB Professional Static capacitance Bluetooth keyboard 、Tesla V100 GPU Allica 、 Baidu Network disk super member And so on ! The uncapped ! There is also a fancy lottery in the live studio waiting for you !

At the end
“ On top of that , It's got to be ; Take one of them , It's down to ; Take it from below , There is no gain ”, lock ViT Punch camp , I hope you can broaden your horizons , Set high goals , Finally, they can achieve their own satisfactory results !
边栏推荐
- A ZABBIX self discovery script (shell Basics)
- Guanghetong and anti international bring 5g R16 powerful performance to the AI edge computing platform based on NVIDIA Jetson Xavier nx
- Codeforces Round #800 (Div. 2)
- 86. (cesium chapter) cesium overlay surface receiving shadow effect (gltf model)
- Lstms in tensorflow_ Cell actual combat
- Resample
- Multipass中文文档-设置驱动
- Datetime data type ---now() gets the current time, datetime() creation date, performs mathematical operations, and to_ Datetime() converts to date type and extracts various parts of date
- Solution to back-off restarting failed container
- 2.< tag-动态规划和常规问题>lt.343. 整数拆分
猜你喜欢

钟珊珊:被爆锤后的工程师会起飞|OneFlow U

Illustration of ONEFLOW's learning rate adjustment strategy

Dbeaver installation and configuration of offline driver

Yolov5 super parameter setting and data enhancement analysis

2022.1.24

Zuul 实现动态路由

Codeforces Round #802 (Div. 2)(A-D)

AD教程系列 | 4 - 创建集成库文件

Second day of deep learning and tensorfow

为什么许多shopify独立站卖家都在用聊天机器人?一分钟读懂行业秘密!
随机推荐
Multipass Chinese document - use multipass service to authorize the client
tensorlow:cifar100_ train
2022.2.11
PHP之一句话木马
钟珊珊:被爆锤后的工程师会起飞|OneFlow U
Datetime data type - min() get the earliest date and date_ Range() creates a date range, timestamp() creates a timestamp, and tz() changes the time zone
date_ Range creation date range freq parameter value table and creation example
2022.2.16
Resample
Cookie and session Basics
Comment enregistrer une image dans une applet Wechat
Numpy index and slice
Introduction to classification data cotegory and properties and methods of common APIs
File upload and security dog
【Unity3D】碰撞体组件Collider
一个从坟墓里爬出的公司
Tensorflow and deep learning day 3
Mise en œuvre du routage dynamique par zuul
Difference between return and yield
Use fill and fill in Matplotlib_ Between fill the blank area between functions