TensorFlow London Meetup: London Meetup: Deep Dive into TensorFlow #28!

Welcome to TensorFlow London Meetup!

PLEASE REGISTER HERE - https://www.eventbrite.com/e/tensorflow-london-meetup-deep-dive-into-tensorflow-28-tickets-92360222913

*UPDATE: We've reached maximum capacity for this event and RSVP's are now closed! Sorry for any inconvenience.

AGENDA:

6:00 - Doors open. Networking. Drinks & Pizza
6:45 - Introduction
7:00 - 7:25 Talk 1: Computer vision in football with TensorFlow by Christopher Evi-Parker, Computer Vision Consultant at Cepco Sports Vision
7:25 - 7:30 Q&A
7:30 - 7:55 Talk 2: Build your own agent with TensorFlow 2.0. for StarCraft II by Gema Parreño Piqueras
7:55 - 8.00 Q&A

__________________________________________________________________________________

TALK DETAILS:

Talk #1
Title: Computer vision in football with TensorFlow

Speaker: Christopher Evi-Parker, Computer Vision Consultant at Cepco Sports Vision

Bio: Christopher Evi-Parker is a computer vision engineer. He started his working life as a software developer optimising operational weather, flooding, and climate models on super computers back when optimising memory access was still a new idea. He then moved to London and to the ‘darkside’, working in finance - from retail payments systems to investment banking. A few years on, and he was itching to find something new to sink his teeth into when he discovered the new fangled field of computer vision and ML (and is an unashamed fanboy of Andrew Ng). As a big sports fan, he wanted to find a way of combining ML and sports - in particular, football. After drumming up interest from what would be future clients at various meetups, he founded his own consultancy business, working with both a startup and an established player in the betting markets, using computer vision methods. Having delivered on key projects, he recently decided to take a step back from consulting and join one of the major players in the field, where, as a computer vision engineer he now has the opportunity and resources to develop and execute on his ideas.

Abstract:This talk is just as much about his recent journey as a computer vision engineer as it is about computer vision in football. Chris will be talking about tools and analysis he found useful, some pitfalls he hit along the way and how best to avoid them.

He will also talk about a specific project he completed with Smartodds, a sports statistics research and sports modelling services company. The objective of this work was to extract football data from archived broadcast videos using computer vision techniques taken from both recent deep ML methods and ‘old school’ multi-view geometry. They successfully extracted player positions data from these videos, which were then used to derive further metrics in combination with other data sources.

Talk #2

Title: Build your own agent with TensorFlow 2.0. for StarCraft II

Speaker: Gema Parreño Piqueras

Bio : Gema has had her own Video game startup and worked as a data scientist in the financial world. She attended Blizzcon workshop about pysc2 API and has produced open source material for the StarCraft II community.

Abstract: This talk gives a general overview about StarCraft II pysc2 API and shows results with A3C in customized mini-games. Besides, it digs into implementation of a variation of DQN: Double Q Learning with dueling architecture using TensorFlow 2.0.