TensorFlow Dev Summit 2018 viewing party

“Problems that were impossible are now possible”

On 22th May 2018 at Fintech District, Data Science Milan in collaboration with Google and BCG Italy have gathered its community to view together some of the main talks of TensorFlow Dev Summit 2018 from last March in Mountain View (CA). TensorFlow is an open source library for machine learning started as a library for deep learning and neural networks. Now is a machine learning platform collecting many algorithms with the goal to make easier to use them.


TensorFlow represents a revolution in the field of machine learning and helps to build artificial intelligence applications; problems that were impossible to solve before are now solved using this technology.

TensorFlow has added value to many different areas such as astronomy (discovering a new planet), healthcare (help to asses a person’s risk for cardiovascular diseases looking at scans of the human eye), aviation (predict the trajectory of a flight) and many others applications.

TensorFlow is at the forefront of machine learning, making it all possible; it is a platform that can solve challenging problems for all of us, it is powerful, scalable and its popularity has grown in the last two years with several innovations, such as TensorFlow Hub, a library to help developers share and reuse models.


Look at the video

“Machine Learning in JavaScript”

TensorFlow.js is an open-source library you can use to define, train, and run machine learning models entirely in the browser, using JavaScript and a high-level layers API.

It uses TensorFlow Playground, an in-browser visualization of a small neural network and it shows in real time all the internals of the network that it’s training.

The browser has become a development environment where you can share the things you build, with anyone with just a link; people that open your app don’t have to install any drivers and can give access to the sensors like microphone, camera and accelerometer, making the app highly interactive.

In the livestream Nikhil Thorat and Daniel Smilkov trained a model to control a pac-man game using computer vision and a webcam, entirely in the browser.


Look at the video

“TensorFlow Lite”

TensorFlow Lite is a lightweight library and tools for doing machine learning on embedded and small platforms, with a different architecture from the one TensorFlow uses: there is an interpreter which runs on-device,  there are a set of optimized kernels and then there are interfaces you can use to take advantage of hardware acceleration when it is available.

It’s cross-platform, it supports Android and iOS and also have support for Raspberry Pi and most of other devices which are running Linux.

In the workflow you take a trained TensorFlow model and then you convert it to the TensorFlow Lite format using a converter then you update your apps to invoke the interpreter using the Java or C++ APIs.



Look at the video 

“Applied AI at The Coca-Cola Company”

TensorFlow has granted to update Coca-Cola North America loyalty marketing programs into a mobile-web platform. The pipeline starts from a pin-code recognition from a cap by OCR (Optical Character Recognition) system to apply a CNNs (Convolutional Neural Networks) combined with TensorFlow to train and predict strings from images that contain small character sets with lots of variance. An active learning system with user interface by a feedback loop allows the model to gradually improve by returning correct predictions to the training pipeline.

Look at the video

Reviewed by Fabio Concina

Author: Claudio Giancaterino

Actuary & Data Science Enthusiast

Follow up