Route planner data for end-to-end driving dataset
Organization
Autonomous Driving Lab
Abstract
End-to-end driving models using just camera input can only do lane following, because they are missing the additional information whether they should turn left or right on the intersection. One way to overcome this limitation is to provide navigational instructions as an input to the end-to-end driving network. These can be represented as navigational commands (e.g. “Turn right after 200 meters”) or route planner screenshots. The task in this project is to add navigational instructions to the end-to-end driving datasets.
The project consists of:
1. Implementing rendering of route planner screenshots using some navigational API, e.g. Mapbox.
2. Collecting sensor data together with navigational instructions, i.e. navigational 3. commands and route planner screenshots.
3. Postprocessing and cleaning the dataset.
4. Training baseline end-to-end driving model conditioned on navigational instructions.
The project consists of:
1. Implementing rendering of route planner screenshots using some navigational API, e.g. Mapbox.
2. Collecting sensor data together with navigational instructions, i.e. navigational 3. commands and route planner screenshots.
3. Postprocessing and cleaning the dataset.
4. Training baseline end-to-end driving model conditioned on navigational instructions.
Graduation Theses defence year
2020-2021
Supervisor
Tambet Matiisen
Spoken language (s)
Estonian, English
Requirements for candidates
Level
Masters
Keywords
Application of contact
Name
Tambet Matiisen
Phone
5286457
E-mail