Google, Facebook, and Microsoft headline this week’s breakthrough-packed AVBytes weekly round-up! And we have a feeling it’s not the last time we’ll be saying that. :)
Highlights from the past week: Google’s Free ‘What-If Tool’, Microsoft’s Open Source Python AutoML Toolkit, MIT’s Computer Vision breakthrough, and more resources and articles below!
Like what you read? Then go ahead and subscribe here to get all the latest developments and resources delivered straight to your inbox daily!

- Google has Open Sourced the Amazing ‘What If Tool’ to Perform Code-Free Visual ML Experiments: Google has launched the ‘What-If Tool’ which lets you analyze your model without writing any code! The visual interface makes performing experiments like hyperparameter tuning, and model result analysis, far more intuitive than ever before. And it’s open source! This is sure to become an integral part of most data scientists’ toolbox.

- Rosetta — How Facebook uses Machine Learning to Process Text in Billions of Images: How does Facebook use machine learning to extract text from images? In this article, we look at the really cool two-step text extraction model Facebook uses to process text in over a billion images every day. A must-read for all NLP enthusiasts!

- Tired of Debugging Code? Facebook’s SapFix Tool Automates the Entire Process: Are you exhausted of going back and looking for bugs in your code? As a data scientist or engineer, you would rather focus on building things and getting them into production, right? So Facebook has developed a tool that automatically fixes bugs in your code and even send a recommendation!

- Perform Automated Machine Learning for Free with Microsoft’s Open Source Python Toolkit: Want to try automated machine learning for free? Microsoft has open sourced a Python toolkit for automating tasks like hyperparameter tuning. In this article we look at the toolkit in a bit more detail and how you can install and use it.

- Jupytext lets you use Jupyter Notebooks as Julia, Python and R Scripts or Markdown documents: Do you prefer editing your code in a script? Then Jupytext is just the tool for you! You can now edit Jupyter notebooks as plain Python, R or Julia scripts, and convert notebooks to and from R markdown documents as well. Check out more details, including installation steps, in this article.

- MIT Open Sources Computer Vision Model that Teaches Itself Object Detection in 45 Minutes (with GitHub codes): Can a computer vision model teach itself to detect objects and manipulate them? It can now! MIT researcher’s have designed such a model using self-supervised learning, and the system can train itself in less than an hour. The code to implement this is available on GitHub as well. A must-read!

The above AVBytes were published from 10th to 16th September, 2018.
Source: Deep Learning on Medium