MicroVision - Utveckling, Testning, och Demonstration av ett Stödsystem I Realtid för Förare av Elektriska Fordon

Purpose and goal:

The project aims to develop a real-time, camera-based rider assistance safety system for e-scooters and e-bikes. The system will consist of low-cost sensors to increase availability. The purpose is to promote safer interactions between road users in urban environments. The development involves gathering and annotation of image data focused on micro-mobility vehicles.

Expected results and effects:

A collected object detection dataset will be shared together with model weights for an image classifier. We believe that the sharing of research artifacts will promote further research into safety and sustainability of small electric vehicles. Our findings will be documented in a scientific paper. In addition, we aim to publish any novelties related to depth estimation in a separate scientific paper, to promote further research.

Approach and implementation:

The first steps of the development are data collection and annotation, along with training of machine learning models. We aim to repurpose state-of-the-art models for object detection as well as depth estimation. The models will eventually be integrated into a system and deployed to hardware for real-time testing on the target vehicle types (e-scooter and e-bike). Finally, the functionality of the system will be demonstrated in a lab environment.