The PaddlePaddle high-T group will lead you to learn "Generative Adversarial Networks" , theoretical analysis + code practice, and will take you to gradually master the core ideas of GAN, common model context and application codes; truly understand GAN, use GAN, and make good use of GAN.
Download and install command ## CPU version installation command pip install -f https://paddlepaddle.org.cn/pip / oschina /cpu paddlepaddle ## GPU version installation command pip install -f https://paddlepaddle.org.cn/pip / oschina /gpu paddlepaddle-gpu
Start time: The camp officially starts at 8:30 pm on Thursday, 1 hour per night, 7 days of lectures, and the whole course is live! ｜April 15 to April 22
Course official group (watch the live broadcast, exchange learning): search the group number 651940985 or scan the code at the end of the article to join the group~ ｜7 high-level teaching assistants with DL codes over 5 years live in the group to answer questions online
— Why learn Generative Adversarial Networks? —
Today, as the CV algorithm is getting introverted, GAN (Generative Adversarial Networks, Generative Adversarial Networks ) continues to attract everyone's attention. It is true that GAN has outstanding performance in action transfer, image super-score , etc., and can be regarded as the top of deep learning cultural creativity .
But in fact, the contribution of GAN is not only reflected in entertainment, but also has many applications in academia and industry . For example , in the field of biology and medicine: medical data is relatively small and difficult to obtain, so using GAN to expand data is a good solution. It is estimated that in 2016 and 2017 alone, more than 400 papers were published in major medical imaging conferences and journals such as MICCAI and MIDL[1][2][3]. In addition , based on the characteristics of GAN confrontation generation, it can also complete the work of generating images from text, high-quality translation, and composing music . More and more fields have begun to pay attention to GAN and make good use of GAN.
Since GAN was proposed by Ian Goodfellow in 2014, it has developed into one of the most promising methods for unsupervised learning on complex distributions in recent years. Hailed by "father of convolution" Yann LeCun as "one of the most interesting ideas in computer science of the past decade". In Baidu Academic , the content of GAN has reached as many as 100,000!
I have to say, GANs are really interesting and useful. But why is the GAN direction algorithm post so high-paying but unable to recruit people? It is true that the model structure of GAN is relatively special, and the model measurement methods are also different. It is not easy to rely on self-study.
We hope that the developers of Baidu Flying Paddle will not only demonstrate with GAN, but also understand the principle, model context and specific code of GAN. Therefore, while providing algorithms and other support for users, we also prepared the open and free "Generative Adversarial Network 7-Day Punch Camp" , and open-sourced all the model codes involved in the course. Maybe you need it in front of the screen~
-Course Introduction-
-Syllabus- _
- Day1_GAN basic concept and application introduction
Detailed explanation of the core idea of GAN, WGAN code, etc.
- Technological evolution and generative application of Day2_GAN
Lectures on DCGAN, LS-GAN, stylegan
- Day3_Image translation and cartoon drawing application
Lecturer on CGAN, pix2pix, cyclegan, Photo2Cartoon
- Day4_Super resolution and old video repair
Lectures on SRGAN, ESRGAN, EDVR and other models
- Day5_Theory and Practice of Movement Transfer
Explain the FirstOrderMotion model in detail
- Day6_Wav2lip Lip Synthesis Theory｜Open Source Construction
Big homework: code question - self-selected generative model to achieve super-score
- Day7_Camp _
Big homework analysis and development improvement｜Contest question guidance
-Teaching Team-
-Teaching Assistant Team-
-Learning Motivation-
1st prize: HHKB Professional electrostatic capacitance Bluetooth keyboard
Second prize 2: Kindle paperwhite e-reader
Six third prizes: Xiaodu wireless smart earphones
20 Excellence Awards: Optional paper books_"Generative Deep Learning", "Deep Learning", "Zero Basic Practice Deep Learning"
The overall score is determined by your code and open source spirit! Students who reach the completion level will receive a certificate of completion officially issued by Baidu Fei Pao + 100-hour GPU computing power card, which will contribute to students who are eager to improve!
At the end of the writing, the R&D brother carefully prepared the preview documents, objective questions and code assignments with progressive difficulty. Don't be afraid that you can't learn, try it first and give yourself more opportunities~
>> Visit PaddlePaddle official website to learn more about it .