Blog: Aerobotics Another Day In Tech | April Recap
In April our Data Scientists have made headlines and kicked off a series of exciting networking events. Let’s look back on what happened in tech this past month at Aerobotics.
Deep Learning IndabaX 2019
What is it?
The Deep Learning IndabaX is an initiative started in 2017 to help develop and strengthen the artificial intelligence and machine learning community in local African countries.
This year it was locally organised and hosted in Durban, South Africa with over 200 attendees and 30 plus speakers over 3 days.
Aerobotics at the IndabaX
This year Aerobotics sent three representatives from the company and sponsored a hackathon for the attendees to participate in.
One of our representatives, Head of Data Science Stuart Reid gave a talk on his current Master’s research; Garth Wasson and Scott Glover put together a hackathon notebook, and Scott Glover took home an award.
Stuart Reid’s Talk
The title of his talk was “The Temporal Turing Test: Sampling Real-Valued Multivariate Time Series From Deep Generative Models”.
This entails looking at sampling real-world medical data (i.e. heartbeats) and generating synthetic representations using generative adversarial models (GAN’s).
Learn more about Stuart and his position at Aerobotics:
Garth Wasson and Scott Glover, two of our Data Scientists at Aerobotics put together a hackathon notebook with the core idea of exploring GAN’s for tree segmentation.
Pride swept across Aerobotics HQ after our Junior Data Scientist, Scott Glover took on the IBM hackathon challenge and won a joint first place!
Glover had to classify whether a genomic sequence was Zootonic or non-Zootonic. To this end, he implemented a machine learning model which focused on feature engineering and data augmentation techniques.
Watch our highlight reel from the Deep Learning IndabaX:
Aerobotics’ Computer Vision Journal Club
On 10 April we held the first of our Computer Vision Journal Club series at Aerobotics HQ!
Head of Data at Aerobotics, Dr Michael Malahe, spoke to a room full of members of the machine learning community.
Our guests were treated to pizza and refreshments — and of course, inside information on how Aerobotics works with computer vision.
A bit about computer vision, from Dr Michael Malahe:
“In computer vision, our goal is to automatically turn electromagnetic signals, like light, into understanding. We know we’ve understood well when we’re able to use our understanding to achieve our goals. For example, if our goal is autonomous driving, we can say we’ve understood an image of a street if we’re able to navigate it safely.
The field of computer vision has grown enormously since deep learning techniques became feasible in the early 2010s. Since September 2012 there have been over 15 000 computer vision publications and preprints tracked on arXiv, the leading open catalogue for this kind of research. The most effective work has involved a high degree of collaboration, with researchers engaging across a wide variety of applications.”
Dr Malahe’s talk
Dr Malahe’s talk introduced listeners to common learning tasks including image classification, object detection, semantic segmentation, image registration, instance segmentation, and multi-modal problems.
Dr Michael Malahe introduces Aerobotics’ Computer Vision Journal Club event series.
Aerobotics and computer vision
At Aerobotics, we work on computer vision for applications like the early detection of agricultural pests and diseases and improving yield for farmers.
At these Computer Vision Journal Club meetings, our aim is to bring a stronger degree of connection between computer vision researchers and practitioners in the greater Cape Town area, so that we can collaborate more effectively, and grow our capabilities in computer vision together.
Our next Computer Vision Journal Club will take place on 9 May 2019. Stay tuned to our Facebook page for more details.
Keep Up To Date our Tech Team!
Our Data Scientists are on the pulse of the latest in machine learning and AI. The team is always busy either attending or developing innovative events in this space.