Joanna Materzynska, an AI engineer at Twenty Billion Neurons or TwentyBN, gave a talk at PyData Warsaw 2017 in which she presents her company's latest gesture recognition technology:
"In this talk I will introduce a Python-based, deep learning gesture recognition model. The model is deployed on an embedded system, works in real-time and can recognize 25 different hand gestures from a simple RGB webcam stream. it uses a simple laptop camera and allows for building playful user interfaces that are controlled with hand gestures. Lastly I will show a demo of such application." [source]
Unlike other PyData talks which are longer, Joanna's is relatively brief, about 27 minutes long and it goes through the following topics and concepts:
- what was the driving factor to create this recognition system
- how they use the 'crowd' for data collection (they use short-video clips as data inputs)
- the obstacles and challenges faced when using video data in very large datasets
- how they built and what type of network architecture they used
- and more.
They also provide some of their datasets for practitioners who want to build with them. Please watch the full video tutorial below:
To stay in touch with me, follow 
@cristi
@cristiCristi Vlad Self-Experimenter and Author