Video based Activity Recognition

Video based Activity Recognition

Goal:
Activity recognition becomes more important in scenarios like surveillance and human-computer interaction. The challenge lies especially in missing, erroneous and inaccurate annotations during training. The goal should apply machine learning techniques to overcome this problems.

Keywords:

• Multi-Instance / Multi-Label Learning
• Spatiotemporal Key-Points

Requirements:

• Interest in Computer Vision
• Interest in Machine Learning
• C++ and/or Matlab