Dimitris is a Machine Learning Engineer. He holds an Industrial Engineering Diploma from Democritus University and a Masters in Intelligent Information Systems from Aegean University among others. As a member of the University’s Robotics Lab, he co-programmed an awarded anthropoid robot which used CV for object detection and face recognition, NLP for Speech to Text, facial gestures for emotional expression, etc. He also co-developed a GUI for easier graphical programming of robotic and IoT devices. He completed his training on Computer Vision at Fujitsu Labs in 2018. Dimitris took first place in 2017 and 5th in 2018 in the ImageCLEF challenge regarding biomedical images labeling. He has been an instructor in many Deep Learning workshops as a member of Machine Learning Tokyo among others. He works as an ML Engineer at LPixel, a Japanese startup that combines AI with Life Sciences. His research interests span Mathematics, Optimization and Machine Learning.
In this workshop we will show a simple and yet complete approach for the DeepSat (SAT-6) Airborne Dataset hosted by Kaggle. We will see how we can explore the dataset, build the necessary data structures and generators for loading, preprocessing and augmenting the data, build a simple neural network and train it from scratch. The code will be written and explained live and interactively. Make sure you have a Gmail account (for Colab) and a Kaggle account (for Kaggle API) if you want to try it yourself.