Register now After registration you will be able to apply for this opportunity online.
This opportunity is not published. No applications will be accepted.
Deep-Learning Map-based Region of Interest for High-Speed Object Detection
The goal of this work is to develop a Region-of-Interest proposal pipeline for long-range object detection and tracking sensor setups in the railway domain.
Keywords: Region of Interest, Localization, Computer Vision, Robotics, Object Detection, Machine Learning
The ability to detect obstacles lying on the vehicle's path is crucial for the development of autonomous vehicles, such as trains or cars. Detecting objects early enough and therefore at long range is crucial, especially for heavy vehicles like trains.
Wide-angle cameras or dense Lidar scans are not capable of detecting objects at such ranges, however actuated high-focal length cameras and sparse laser scanners can provide results at distances in excess of 1000m. As these are much slower though they require a good prior about where to search for possible objects in the form of a Region-of-Interest (RoI) proposal.
In this project we aim to develop a map-based RoI proposal algorithm by projecting available railway map data into the video stream. In a second step this can be used as a baseline comparison method for evaluating existing machine-learning based RoI algorithms in the railway domain. This can be used to generate further training data in order to improve machine learning based approaches or to create a hybrid approach.
The ability to detect obstacles lying on the vehicle's path is crucial for the development of autonomous vehicles, such as trains or cars. Detecting objects early enough and therefore at long range is crucial, especially for heavy vehicles like trains. Wide-angle cameras or dense Lidar scans are not capable of detecting objects at such ranges, however actuated high-focal length cameras and sparse laser scanners can provide results at distances in excess of 1000m. As these are much slower though they require a good prior about where to search for possible objects in the form of a Region-of-Interest (RoI) proposal. In this project we aim to develop a map-based RoI proposal algorithm by projecting available railway map data into the video stream. In a second step this can be used as a baseline comparison method for evaluating existing machine-learning based RoI algorithms in the railway domain. This can be used to generate further training data in order to improve machine learning based approaches or to create a hybrid approach.
- Review of literature and existing RoI algorithms.
- Development of a map-based RoI proposal pipeline.
- Comparison of map-based and machine learning based approaches.
- Further training of machine learning based approaches and transfer to the railway domain.
- Testing and evaluation using recorded or simulated datasets.
- Review of literature and existing RoI algorithms. - Development of a map-based RoI proposal pipeline. - Comparison of map-based and machine learning based approaches. - Further training of machine learning based approaches and transfer to the railway domain. - Testing and evaluation using recorded or simulated datasets.
- Highly motivated and independent student.
- Interest in machine learning and object detection.
- Programming skills in Python or C++.
- Experience with machine learning frameworks is a plus.
- Enrolled at ETH Zurich.
- Highly motivated and independent student. - Interest in machine learning and object detection. - Programming skills in Python or C++. - Experience with machine learning frameworks is a plus. - Enrolled at ETH Zurich.
If you are interested, please contact Andrei Cramariuc (andrei.cramariuc@mavt.ethz.ch) and Cornelius von Einem (cornelius.voneinem@mavt.ethz.ch), sending your CV, transcript of records and a short paragraph on why you want to work on this project.
If you are interested, please contact Andrei Cramariuc (andrei.cramariuc@mavt.ethz.ch) and Cornelius von Einem (cornelius.voneinem@mavt.ethz.ch), sending your CV, transcript of records and a short paragraph on why you want to work on this project.