You are viewing content from a past/completed QCon

Track: AI Meets the Physical World

Location: Cyril Magnin II

Day of week: Tuesday

Artificial intelligence and machine learning algorithms became an essential part of many online services. Lately many companies started bringing AI to the physical world. Mobile phone apps use more and more on-board machine learning algorithms, cars are becoming self-driving, and drones process their images on the fly!

In this track we explore use-cases of machine learning that touch us directly. We will explore algorithms that work on drones and cars, algorithms that use a human in their loop, use where you live as input, or even augment the physical world. We will look both at the amazing applications AI is applied to, as well as best practices and tools you should use to bring your own application into the physical world.

Track Host: Roland Meertens

Machine Learning Engineer @Autonomous Intelligent Driving

Roland Meertens is Machine Learning Engineer at Autonomous Intelligent Driving. He works on the machine learning side of the perception software stack that will be deployed to the autonomous vehicles that will soon roam urban environments in Germany.

10:40am - 10:50am

The Basics of ROS Applied to Self-Driving Cars

The Robot Operating System, or ROS, is a framework used for robotics around the world. With the emergence of self-driving cars, many companies have helped moved ROS from an academic tool to a core component in the development of their autonomous vehicles. This talk will cover a brief overview of the ROS architecture and how you would begin using it on your self-driving car project.

Anthony Navarro, Product Lead- Robotics and Self-Driving Car @Udacity

11:00am - 11:50am

Self-Racing using Deep Neural Networks: Lap 2

This talk details a team of 17 Udacity Self-Driving Car students as they attempted to apply deep learning algorithms to win an autonomous vehicle race. At the 2017 Self Racing Cars event held at Thunderhill Raceway in California, the team received a car and had two days before the start of the event to work on the car. In this time, we developed a neural network using Keras and Tensorflow which steered the car based on the input from just one front-facing camera in order to navigate all turns on the racetrack. We will discuss the events leading up to the race, development methods used, and future plans including the use of ROS and semantic segmentation.

Anthony Navarro, Product Lead- Robotics and Self-Driving Car @Udacity
Jendrik Jördening, Data Scientist @Nooxit

12:50pm - 1:00pm


  • Quick Overview of the TX2 hardware and TX2 Dev Kit
  • Software that runs on the TX2 Devkit

Examples: Supervised DL for Classification, Segmentation from the camera and RL projects

Dana Sheahen, Robotics @Udacity

1:10pm - 2:00pm

Liquidity Modeling in Real Estate Using Survival Analysis

Hazard / survival modeling is often under-applied given its broad use cases. For example, churn prediction is often posed as a classification problem (did churn or not), when the time component is often given short shrift (when, if ever, did the churn happen?).

We hope to argue that hazard modeling is a better fit for these types of problems; spread general awareness of survival modeling, metrics, and data censoring; and describe how Opendoor uses these models to estimate our holding times for homes and mitigate risk, detailing scalability and other technical challenges we had to overcome.

David Lundgren, Data Scientist @Opendoor
Xinlu Huang, Data Scientist @Opendoor

2:25pm - 2:35pm

Constraints of Building a Modern Drone

This short talk discusses where we are today in 2018 with drone development. The frame and the auto-piloting are actually the easy part. The real issues that remain include issues like processing power, control/communication over large distances, fleet management, regulatory approvals, and situational awareness. Interested in learning more about the real issues with drone development today, this talk will get you up to speed.

Alexander Harmsen, CEO/Founder @iris_automation

2:45pm - 3:35pm

Dive Deep into Building Computer Vision Systems for Self-flying Drones

Every industrial drone in the world needs to see the world the way a pilot does, an obvious, but non-trivial feat of engineering. In this talk the challenge of building a computer vision/AI "sense and avoid" system is discussed, along with the challenges of a constrained cost/size/weight/power design environment. It’ll dive into deep learning training, simulation, and real-world aircraft testing with near-mid-air collisions, in order to keep these drones safe while they get integrated into the national airspace!

Alexander Harmsen, CEO/Founder @iris_automation

4:00pm - 4:10pm

Building a Security System with Image Recognition & an Amazon DeepLens

This quick talk will show you step by step how I build a security system for my house using the upcoming Amazon DeepLens. I'll go over how I built and trained the models, and the steps necessary to get the camera making inferences and sending alerts.

Jeremy Edberg, Cofounder @CloudNative

4:20pm - 5:10pm

Deep Learning for Science

How is deep learning affecting the world of science? In this talk, Prabhat (Data and Analytics team lead at NERSC, Berkeley Lab) discusses machine learning's impact on climatology, astronomy, cosmology, neuroscience, genomics, and high-energy physics. The talk will conclude with a list of open challenges in applied Deep Learning, and the future of AI in powering scientific discoveries.

Prabhat , Data and Analytics Group Lead @NERSC

2019 Tracks

  • ML in Action

    Applied track demonstrating how to train, score, and handle common machine learning use cases, including heavy concentration in the space of security and fraud

  • Deep Learning in Practice

    Deep learning use cases around edge computing, deep learning for search, explainability, fairness, and perception.

  • Handling Sequential Data Like an Expert / ML Applied to Operations

    Discussing the complexities of time (half track) and Machine Learning in the data center (half track). Exploring topics from hyper loglog to predictive auto-scaling in each of two half-day tracks.

    Half-day tracks