Forskningsamanuens inom förståelse av avvikande beteenden i trafik
Core
Prepare video datasets of traffic incidents, collect and annotate public videos, and support fine-tuning and evaluation of a vision-language model for detecting at-risk events involving vulnerable people.
Role type
Research assistant (computer vision and multimodal AI)
Builds
Annotated video datasets and evaluated AI models for traffic safety
Domain
Transportation safety + computer vision
Deliverable
production ML models
Required skills
Python, machine learning, computer vision, video processing, data preparation, annotation, event detection, localization
Preferred skills
vision-language models, PyTorch, fine-tuning models, evaluation of AI systems
Responsibilities
Prepare video datasets of traffic accidents and cases, collect and annotate publicly available videos, create labels for event time boundaries, spatial positions, causes, and severity, prepare question-and-answer data, support fine-tuning and evaluation of a vision-language model, analyze event detection and localization in time and space, document experiments and research results
