Live Object and Keypoint Detection

Institute for Machine Learning, JKU Linz (AT)

In order to identify the locations and types of objects in an image, computer vision models based on deep learning can be used. Here, we show two examples:

  1. Object detection: An artificial neural network was trained on 80,000 images. It can recognize 80 different classes of objects such as people, phones, and bananas in real time.
  2. Keypoint detection: This is a real-time multi-person system, which is able to jointly detect hand, facial, and foot keypoints of humans. This model was trained on 250,000 annotated instances.

Here, with a webcam connected to a powerful computer, you can experience the performance of these deep learning algorithms in a live environment.

Institute for Machine Learning, Johannes Kepler University, AUT (https://www.jku.at/iml)
Facebook AI Research, USA (https://github.com/facebookresearch/Detectron)
Carnegie Mellon University, USA (https://github.com/CMU-Perceptual-Computing-Lab/openpose)