Label-efficient learning of LiDAR-based perception models for autonomous driving

• Type: Thesis
• Date: 2021-07-22

Authors

Bernardo Magina Madureira Palha de Araújo (Author) FEUP

Thesis

Label-efficient learning of LiDAR-based perception models for autonomous driving

Master's Degree

Dissertation

Scientific classification

FOS: Engineering and technology > Electrical engineering, Electronic engineering, Information engineering

Other information

Resumo (PT): Deep learning on 3D LiDAR point clouds is in its infancy stages, with room to grow and improve, especially in the context of automated driving systems. A considerable amount of research has been pointed at this particular application very lately as a means to boost the performance and reliability of self-driving cars. However, the quantity of data needed to supervise perception point cloud-based models is extremely large and costly to annotate. This thesis studies, evaluates and compares state-of-the-art detection networks and label efficient learning techniques, shedding some light on how to train perception models on point clouds with less annotated data.

Language: English

No. of pages: 70