Collaborative Visual Place Recognition

Target Audience: Bachelor's and Master's Students

Topic:

Visual place recognition (VPR) consists in the ability to recognize previously seen places. Typically using images as input, this technique is a rapidly growing field of research and plays an essential role for localization and mapping in robotics, augmented and virtual reality.


Project:

The proposed project aims to investigate efficient methods, whether using classical or AI approaches for use in collaborative mapping with resource-constrained devices such as small/portable computers. The project work stages comprise the following ordered-topics:

  • A survey on literature, classical x AI methods
  • Implementation and benchmarking of methods on specific resource-limited devices
  • Aim to provide contributions or improvements to existing methods


Requirements:

  • Interest for the topic (computer vision, robotics, XR)
  • Knowledge or interest to rapidly learn:  C++, Linux, Python and AI frameworks (Pytorch/TensorFlow)
  • Pro-active and motivated attitude

If you are interested, please contact Carlos Pinheiro.