Master Thesis - Deep Learning for 3D Point Clouds

Deep learning has revolutionized many fields, especially computer vision, where you get state of the art results in areas such as classification, object detection and segmentation. 

Most focus has been on using images as input to deep neural networks but in some applications only 2D information is not enough.

SICK has different kind of sensors which produce 3D information and this thesis is about finding known objects within a 3D point cloud with deep learning, for example using object detection or instance segmentation.

Aspects that might be considered are:

• Deformable objects
• Cluttered scenes
• Generalization from synthetic data
• Incorporation of temporal data
• Efficiency during inference

Prerequisites

• Machine learning
• Computer vision
• Programming in Python
• Deep learning framework such as TensorFlow

Contact

For more information, contact R&D Manager Andreas Wrangsjö, andreas.wrangsjo@sick.se

We look forward to your application before  the 18th of October!