Hey, how's it going, everyone! In this video we take a look at a really cool GitHub repository I found that allows us to easily train a Keras neural network to recognize our own custom hand gestures, and runs flawlessly on a CPU. But there's more: I actually guide you through the entire process and also explain the logic of how everything works behind the scenes. We will use Mediapipe to extract hand landmarks (which runs a pipeline of TensorFlow neural nets under the hood), allowing us to use relatively small training datasets for training our new custom hand gestures.

GitHub repository (translated into English): https://github.com/kinivi/hand-gesture-recognition-mediapipe
Original GitHub repository (in Japanese) https://github.com/Kazuhito00/hand-gesture-recognition-using-mediapipe
A series of videos I made about OpenCV: https://www.youtube.com/playlist?listPLZBN9cDu0MSlHMjJughb11ydICWbA8OBe
Here's the same repository, but in React and nextJS: https://github.com/TomasGonzalez/hand-gesture-recognition-using-mediapipe-in-react

Follow me on GitHub: https://github.com/ivangrov
Contact me on Twitter: https://twitter.com/Ivangrov (fun updates on more videos I make there)
And on LinkedIn: https://www.linkedin.com/in/ivangrov/

Timestamps
00:00 Introduction
01:14 What will be in the video?
04:00 Top-level overview of the hand gesture recognition approach we will use
14:56 Google's Mediapipe framework Python API
17:50 Hand Gesture Recognition GitHub repository we will use
20:58 Changing the repo code to enable multi-hand detection
25:23 Code explanation
32:46 Explanation of the hand landmark preprocessing algorithm
48:05 Disable point history classification
50:43 Training: Hand gesture dataset
53:30 Training: Adding a new hand gesture to the existing one
01:02:43 Training: Retraining the model with brand new hand gestures
01:09:51 Outro

Contact me directly: [email protected]

It would mean a lot to me if you decided to support me and the channel)

You may want to consider watching ads that appear with the videos

Making these videos takes a lot of time and effort, so if you decide to support me, please don't hesitate to contact me as I would like to thank you personally!

Thanks for looking!

Please take the opportunity to connect and share this video with your friends and family if you find it helpful.