The goal of this project is to construct a real-time robust semantic scene representation that can help robots understand and navigate in their environments. While most existing works focus on indoor representations for ground robots, we will be focusing on the representation's utility for outdoor unstructured environments. The representation should be useful for both a ground and aerial robot.
If you are interested, before reaching out please read the following papers:
- Strong coding skills in python and particularly PyTorch.
- Knowledge of unsupervised learning (Dimensionality reduction techniques, clustering algorithms...etc.) in addition to deep learning.
- Ability to read, understand, and replicate results of papers.
- Proactive.