開放空間導覽 Guide to Open Spaces|Benson / Shirley|Gather|PyCon APAC 2022
開放空間(Open Spaces)是自助式、聚會形式的活動,與大會議程同時進行。開放空間的主題是由大會與會者「當場」計畫的。開放空間讓你能用任何你喜歡的方式定義、組織、規劃你自己的小聚。也歡迎你揪其他好夥伴一起來參與、分享、聊天、交朋友!對於曾經參加過過往 PyCon Taiwan 或是其他研討會的會眾,開放空間的概念也許讓你覺得似曾相識。沒有錯,開放空間基本上很類似 BoF ,不過「開放空間 / Open Spaces」對於第一次參加的會眾來說更容易理解。
Open Spaces are self-organizing meetup events that happen simultaneously with the main conference. Open Spaces are organized during the period of PyCon by attendees, which provides a way for you to define, organize, plain out the meetup as you preferred. Enjoy the Open Space while making new friends, chat about any topic you’d like at the same time.For those who have participated in previous PyCon Taiwan, Open Spaces may sound familiar to you. Yes! it’s basically the good old BoF (Birds of a Feather), but with a name that is more comprehensive to new conference attendees.
PyCon APAC 2022|一般演講 Talks|國泰金控 Cathay Financial Holdings / 美光科技 Micron 冠名贊助
✏️ 共筆 Note:https://hackmd.io/@pycontw/S1tgw6Xki
?? Slido:https://app.sli.do/event/1BdbMfQK1qZCe1HRbbWyd4
? 語言 Language:英文 English
? 層級 Level:中階 Intermediate
? 分類 Category:圖像處理 Graphics
? 摘要 Abstract ?
There are many ways to make your photos and videos better and clear, such as super resolution, image colorization and video stabilization. One of the fundamental techniques is image denoising, i.e., to reduce noise from images. In the talk, we will firstly introduce the background and its related applications about image denoising. Then, a classic Deep Learning method: DnCNN will be taken as an example, to describe its main concepts and implementation in PyTorch. Finally, we will show a practical use case for old film restoration.
? 說明 Description ?
Challenges for Image Denoising
In the past, many traditional image denoising algorithms assume that the noise is under some well-known distributions, such as Guassian distribution. However, the distribution is unknown in real scenarios, which usually makes two kinds of results: the image is still noisy or the image becomes blurred. We’ll illustrate the situation (with examples) and possible ways to solve the problem by using Deep Learning. The main difference between traditional methods and Deep Learning methods is the latter apply modern Neural Network architectures and learn how to denoise from data.
Related Applications
We can apply the same image denoising technique to other tasks. For instance, DnCNN can directly solve the problem of super resolution (a task to enlarge the resolution of an image). In general, many applications about image restoration are well fit to the algorithms of image denoising.
DnCNN and other Deep Learning Methods
We’ll show the main components of a DnCNN model: Convolutional layer, Batch Normalization, ReLU, but details are ignored because materials about complicated mathematics are NOT the targets. Such concepts help when people try to build the network architecture. We will also introduce Residual Learning, an important technique used by DnCNN. Then, we will show how to train a DnCNN model by PyTorch, a popular Python based Deep Learning framework. If time permits, we may also compare DnCNN with the other popular solution: U-Net, an Auto-Encoder based method.
Applying to Old Film Restoration
The pain points of the old film restoration industry will be discussed. We'll tell the audience how to fit the problems of old film restoration to image denoising and use related techniques to solve the problems. Also, some difficulties we encountered in the processing of solving the problem, e.g., the trade-off between restoration capability and artifacts generation, will also be shared. The solutions include: the choice of Deep Learning models, the training method, data pre-processing and post-processing.
? 講者介紹 About Speaker - 蘇嘉冠 (Jia-Kuan, Su) ?
J.K. is an AI Researcher and Developer in startup, he is always willing to share and discuss AI techniques with people. He is also an advocate of Unconditional Basic Income (UBI) in Taiwan.
#pycontw #pyconapac2022 #python #imageprocessing #imagedenoising #videostabilization #pytorch
Follow “PyCon Taiwan”
⭐️ Official Website: https://tw.pycon.org
⭐️ Facebook: https://www.facebook.com/pycontw
⭐️ Instagram: https://www.instagram.com/pycontw
⭐️ Twitter: https://twitter.com/PyConTW
⭐️ LinkedIn: https://www.linkedin.com/company/pycontw
⭐️ Blogger: https://pycontw.blogspot.com
...
https://www.youtube.com/watch?v=ddi58EcZPxw
Day 2, 13:00-13:45
Abstract
過往在使用Python開發API的時候,基於WSGI架構下,運算中遇到IO bound時,可以透過多線程去處理,但遇到cpu bound時,多線程並不是一個好的選擇,因為線程會受限於GIL,並不會有效提升效能,而當遇到運算請求量較大時,為了因應大量的運算請求,可以透過多進程來彌補多線程的不足,但同時也需要承擔過多的資源消耗以及考慮Inter-Process Communication overhead。 而這些問題在FastAPI得到了救贖,FastAPI是一個建立在ASGI架構下的Web框架,以Python所提出的非同步概念為基礎,透過Coroutine的方式,去提高CPU運算效率,去改善多線程、多進程對於cpu bound遇到的問題。 本次演講主要透過房屋估價模型服務程式碼的解說,分享如何從過去使用的Flask框架,轉換為FastAPI框架。透過非同步的設計,即使有GIL的限制,仍舊能夠透過單線程去達到類似多進程的運算效能,減少開啟多進程造成的資源耗費過多。此外,透過程式碼,說明這兩個框架開發上的不同之處,以及轉換過程中曾經遇到的問題,讓想使用FastAPI但還在觀望的人能夠有參考方向,減少未來使用時踩雷的機會!
Description
想透過本次演講告訴大家的內容
ASGI架構是什麼?FastAPI是什麼?是如何應用Async?
藉由實際痛點(mutli-thread、mutli-process issue)案例-房屋估價模型服務分析Flask與FastAPI開發上相異之處,以及這樣的相異之處對開發上的影響
非同步框架帶來好處的同時,使用上所需的注意事項
實際房屋估價模型服務轉換前轉後的數據比較
除了效能以外,FastAPI所帶來的好處
轉換上的心路歷程
本次使用的相關工具
FastAPI
Flask
Slides not uploaded by the speaker.
HackMD: https://hackmd.io/@pycontw/2021/%2F%40pycontw%2Frk6Kg4cfY
Speaker: 陳家丞
Intelligent System Engineer at E.SUN Bank. Like to study python Web Framework, e.g., Flask、FastAPI.
...
https://www.youtube.com/watch?v=5bIJ1SwF2bk
Let's welcome Reuven Lerner from Israel!
Not only did Reuven offer online courses, he also conducted on-site training for big companies around the world. He started his journey with a general tech education, and gradually shifted his direction to Python training. This time he shared his story of how it's like to be a full-time Python trainer and his teaching core value. He also shared a lot about his experience in teaching and helping his student grow.
---
Guess what?
Reuven offer free online courses for those who are starting as a beginner. Go check out his course on his website now!
[Reuven's online course]: https://lerner.co.il/
---
留言告訴我你對這一集的想法: https://open.firstory.me/user/ckmkshy2r2s5a0823k3c1t8uz/comments
Follow “PyCon Taiwan”
⭐️ Official Website: https://tw.pycon.org
⭐️ Facebook: https://www.facebook.com/pycontw
⭐️ Instagram: https://www.instagram.com/pycontw
⭐️ Twitter: https://twitter.com/PyConTW
⭐️ LinkedIn: https://www.linkedin.com/company/pycontw
⭐️ Blogger: https://pycontw.blogspot.com
...
https://www.youtube.com/watch?v=kjiuhMtI_A4
PyCon Taiwan 2016|一般演講 Talks|Geo Processing With Python: How to Convert, Clean, Aggregate and Compress Your Geo-Data for Web
? 摘要 Abstract ?
How to start transforming my shapefiles? What geo formats should I use for web? How to design geo data processing pipeline? What are geojson, topojson, epsg or shapefiles? Join us for this talk and you'll walk away with a better understanding of Geo Processing with Python. We will go through the design process for map-visualization web app from the geo data processing point of view. Starting with the entire design process for the data pipeline including software architecture, used technologies, automation, mapping the data to the geo-data and optimizing the maps for the web. You will have concrete examples on how this approach was applied in a project for the World Bank. You should walk away with the tools and skills to design and implement a geo data pipeline for their own project.
? 關於講者 About Speaker - Juha Suomalainen ?
Originally from Finland, but currently living in Shanghai. I have a degree in telecommunications engineering and have experience on multiple areas including Data Portals, DevOps and Telecommunications. I work as a Software Engineering Director at Wiredcraft. I balance my time between the roles of software architect, tech lead and programmer. My strength is Python but I am used to working in multilanguage systems. I am also the lead engineer for devo.ps (http://devo.ps). Programming has been part of my life since young age. In my free time I like to play sports (any sport!) and learn new technologies. I also like attending and presenting in the local engineering meetups.
#python #pycontw #pycontw2016
Follow “PyCon Taiwan”
⭐️ Official Website: https://tw.pycon.org
⭐️ Facebook: https://www.facebook.com/pycontw
⭐️ Instagram: https://www.instagram.com/pycontw
⭐️ Twitter: https://twitter.com/PyConTW
⭐️ LinkedIn: https://www.linkedin.com/company/pycontw
⭐️ Blogger: https://pycontw.blogspot.com
...
https://www.youtube.com/watch?v=70wOpnEbqSY
Day 3, R0 11:15–12:00
Writing concurrent program is hard; maintaining concurrent program even is a nightmare. Actually, a pattern which helps us to write good concurrent code is available, that is, using “channels” to communicate. This talk will share the channel concept with common libraries, like threading and multiprocessing, to make concurrent code elegant.
撰寫並行程式(concurrent program),例如多執行緒程式等,通常是困難的,而維護並行程式又是一場惡夢。其實有個方法可以寫出好的並行程式,只要使用 channel 做為溝通基礎,就可以寫出好維護的並行程式。這場演講會介紹如何在常見的函數庫,例如 threading、multiprocessing 等,中引入 channel 的概念,讓並行程式變得優雅。
The speaker did not upload his slides.
...
https://www.youtube.com/watch?v=ZT8zxMajRYY
Day 1, 10:10–11:10
Peter will talk about why he thinks Python has come to popularity, its strengths and weaknesses for data analysis, and why an open, accessible language is critical for a future world that will be dominated by machine learning and AI.
Slides not uploaded by the speaker.
Speaker: Peter Wang
Peter Wang has been developing commercial scientific computing and visualization software for over 15 years. He has extensive experience in software design and development across a broad range of areas, including 3D graphics, geophysics, large data simulation and visualization, financial risk modeling, and medical imaging.
Peter’s interests in the fundamentals of vector computing and interactive visualization led him to co-found Anaconda (formerly Continuum Analytics). Peter leads the open source and community innovation group.
As a creator of the PyData community and conferences, he devotes time and energy to growing the Python data science community and advocating and teaching Python at conferences around the world. Peter holds a BA in Physics from Cornell University.
...
https://www.youtube.com/watch?v=p61uHiK7tWo