ADS & SABA Webinar | A-Z in Data Science: Computer Vision

Amsterdam Data Science (ADS) is hosting a new webinar series in collaboration with the Study Associations from the UvA and VU to explore the A-Z in Data Science.

In this webinar we will be exploring Computer Vision, the process of recording and playing back light fragments, when computer and/or machine has sight.


Stevan Rudinac


12:00 Introduction & Welcome
12:05 Talk #1: Adversarial Attacks and Autonomous Cars
12:20 Q&A
12:30 Talk #2: PanorAMS: Identifying and Localizing Objects in Urban Context
12:50 Q&A
13:00 End!

Talk #1 by Benedikt Fuchs

Benedikt Fuchs studied visual computing at the TU Vienna. He’s been working as a Data Scientist at Cloudfight for the past 3 years. He has designed various NLP solutions for real world applications and keeps a radar on various emerging technologies.

Title: What are adversarial attacks and what challenges do they provide to autonomous cars?

While adversarial attacks have not found any real-live uses so far, it has enormous potential to be harmful. One of these potential dangers is the deception of autonomous cars. Benedikt Fuchs will explain how these attacks work and what the consequences of these attacks are.

Talk #2 by Inske Groenen

Inske is a PhD researcher within the Multimedia Analytics Lab Amsterdam at the University of Amsterdam. In her research she focuses on developing cutting-edge Artificial Intelligence techniques that allow us to gain insight and extract information from multiple data sources. She is especially interested in using these techniques to address urban challenges related to social issues, health and sustainability.

In this talk she will give some insight into the field of urban computing. This interdisciplinary field, which connects computer science to more traditional city-related areas, such as environmental sciences, sociology, and economics, focuses on acquiring, integrating and analyzing big data in urban spaces. Specifically, we focus on how multimedia sources available within the urban context can be used to identify and localize common objects within panoramic street level images. To this end, we have developed the PanorAMS framework, which includes a method to get an approximation of which objects are visible in which image, and where, based on urban context information. Following this method, we have developed a new large-scale urban dataset solely from open data sources in a fast and automatic manner. The dataset covers the City of Amsterdam and contains over 15 million annotations of 24 different object categories present in close to 800,000 panoramic images. She will show the significance of this dataset and how it may be used to identify and localize within panoramic street level images.

Zoom details: