Register now After registration you will be able to apply for this opportunity online.
This opportunity is not published. No applications will be accepted.
Machine learning for multicopter based semantic mapping
This project looks at the integration of pixel and object wise semantic classification into an existing 3D dense mapping infrastructure. The system will run online onboard a multicopter as it explores an area.
Here at the ASL we have recently developed a TSDF mapping framework Voxblox (https://www.youtube.com/watch?v=PlqT5zNsvwM). The system provides a reconstruction of the surrounding 3D geometry of an area while operating on board a multicopter.
Recent advances in mobile GPUs and machine learning pipelines now make it possible for a small system to perform image classification in real time. This information can then be fused with the 3D geometry of our mapping framework to provide real time semantic maps of an environment.
The task of this project are to:
- Interface a camera with a jetson tx2 board
- Perform pixel and object classification of the image using deep learning pipelines
- Project this information into a 3D map
- Integrate the hardware into a multicopter system
Here at the ASL we have recently developed a TSDF mapping framework Voxblox (https://www.youtube.com/watch?v=PlqT5zNsvwM). The system provides a reconstruction of the surrounding 3D geometry of an area while operating on board a multicopter.
Recent advances in mobile GPUs and machine learning pipelines now make it possible for a small system to perform image classification in real time. This information can then be fused with the 3D geometry of our mapping framework to provide real time semantic maps of an environment.
The task of this project are to:
- Interface a camera with a jetson tx2 board - Perform pixel and object classification of the image using deep learning pipelines - Project this information into a 3D map - Integrate the hardware into a multicopter system
Not specified
This project can be either a semesters or masters thesis. Prerequists for this project are:
- Familiarity with a linux console
- Experience with ROS
- Programming experience in python and C++
- Previous experience with machine learning libraries or GPU programming is a plus
This project can be either a semesters or masters thesis. Prerequists for this project are:
- Familiarity with a linux console - Experience with ROS - Programming experience in python and C++ - Previous experience with machine learning libraries or GPU programming is a plus
If interested please send your cv to Zac Taylor ztaylor@ethz.ch
If interested please send your cv to Zac Taylor ztaylor@ethz.ch