Usable surface detection on top-view camera data using automatic ground truth generation for binary semantic segmentation
Published in None, 2020
Group Project in Systems, Control and Mechatronics
This project aimed to solve the problem to detect drivable areas of a factory floor, using a workflow consisting of background subtraction to generate a ground truth, that is then used to train a neural network.
The output could potentially be used as a mapping for which a scheduler or robot control system could work upon to avoid collisions and problems with dynamic obstacles.
| Content for this publication | Link |
|---|---|
| Link to the final paper | Final report |
| Link to the fair poster | Poster |
| Link to the video with potential drawbacks | Video |
Technologies used: Intel Realsense Camera, OpenCV, GrabCut, PyTorch
Publication: Download the final publication here!
Presentation/Demonstration: Link to the corresponding presentation
Recommended citation:
Rauh, Lukas. (2020). Usable surface detection on top-view camera data. (MPSYS Design projects 2019/2020, Chalmers University of Technology, Gotheburg, Sweden)