Building a Gesture Recognition System using Deep Learning (video)

Here is a talk by Joanna Materzynska, AI engineer at TwentyBN, which was recorded at PyData Warsaw 2017. In the talk, Joanna presents TwentyBN’s latest gesture recognition technology.

In her PyData talk, Joanna explains the process of building state-of-the-art gesture technology and shares some of the lessons we learned, including:

  • Motivation for building a RGB video-based gesture recognition system
  • How we conduct a large-scale crowd-acting operation to collect hundreds of thousands short video clips
  • Challenges with training neural networks on massive video datasets
  • The development of a network architecture that allows for classifications of video clips solely with RGB input frames
  • The iterations necessary to make the neural network run in real-time on embedded devices
  • Lastly, the discovery and development of playful gesture-based applications.

You can find some of TwentyBN’s datasets here. To learn more about TwentyBN, visit our website or contact us.

TwentyBN Jester Dataset: Recognize Human Hand Gestures